Nomination: Best Hardware, High School

Overview

Remedix is a compact, AI‑powered vision system with Internet of Things (IoT) connectivity, built for automated detection, data collection, and real‑time environmental monitoring. In this project, it serves as an intelligent biosensing platform that tracks chromoprotein signals produced by genetically modified (E. coli) cultures encapsulated in alginate beads. Together, these elements form a portable biosensor network for detecting pharmaceutical waste and antibiotic residues in water.

3D model of the Remedix modular system

The engineering of the machine body of Remedix

The testing of the components of Remedix

Concept and Design of Remedix

1. Why Use Alginate Beads?

By reviewing past biosensor designs, we identified two major challenges: (1) the unclear color change produced by GM E. coli, and (2) the inconvenience and biosafety concerns associated with carrying or handling live bacterial cultures, which make the process less appealing to non‑expert users and stakeholders.

Remedix utilizes spherical calcium alginate beads (≈ 3 mm in diameter) containing micropores that allow the diffusion of enzymes and small molecules while preventing microbial leakage (Klein, Stock & Vorlop, 1983). This structure permits analytes such as tetracycline and salicylate to diffuse into the alginate matrix for detection while securely retaining the bacteria. By concentrating E. coli cells within these beads, Remedix enhances colorimetric sensitivity and user convenience, generating stronger and clearer color responses without the need to handle or culture free bacterial suspensions.

Research further supports the effectiveness of this approach. E. coli immobilized in calcium alginate has demonstrated stability and sensitivity for up to 20 days (Wasito et al., 2019). In addition, alginate’s biocompatibility, scalability, and low cost (below USD 100 per kg) make it an ideal material for sustainable and deployable biosensors (Lee & Mooney, 2012). Integrating these properties, Remedix emerges as a robust, affordable, and field‑ready biosensor tailored for the precise detection of pharmaceutical residues such as tetracycline and salicylate.

Reference:

• Klein, J., Stock, J., & Vorlop, K.‑D. (1983). Pore size and properties of spherical Ca‑alginate biocatalysts. European Journal of Applied Microbiology and Biotechnology, 18, 86–91. https://doi.org/10.1007/BF00500829
• Lee, K. Y., & Mooney, D. J. (2012). Alginate: Properties and biomedical applications. Progress in Polymer Science, 37(1), 106–126. https://doi.org/10.1016/j.progpolymsci.2011.06.003
• Wasito, H., Fatoni, A., Hermawan, D., & Susilowati, S. S. (2019). Immobilized bacterial biosensor for rapid and effective monitoring of acute toxicity in water. Ecotoxicology and Environmental Safety, 169, 774–780. https://doi.org/10.1016/j.ecoenv.2018.11.141

2. Optimization of Alginate and Calcium Chloride Concentrations for Bead Formation

After selecting alginate beads as the encapsulation material, we conducted a series of experiments to optimize the mass ratios of sodium alginate and calcium chloride to water for stable bead formation. Various combinations were tested to evaluate membrane strength and permeability, as summarized below.

Conc. of Sodium Alginate to Conc. of Calcium Chloride	Strength of beads formed after 1 min
1:30 to 1:50	+
1:30 to 1:10	++++
1:60 to 1:50	++
1:60 to 1:10	+++

(“+” indicates relative membrane strength)

During bead formation, the sodium alginate solution is mixed with the bacterial culture in a 1:1 ratio, then added dropwise into the calcium chloride solution to initiate crosslinking. From our experiments, we determined that the most favorable mass ratios are 1:60 (sodium alginate : water) and 1:10 (calcium chloride : water). A more concentrated alginate solution (such as 1:30) was found to be too viscous, resulting in poor droplet formation and irregular beads.

We tested varying concentrations of sodium alginate and calcium chloride to determine the optimal conditions for forming stable bacterial beads.

3. Functional Test of Alginate Beads

During the developmental stage, our tetracycline‑ and salicylate‑sensitive GM E. coli strains had not yet been successfully cloned, so as an alternative for proof of concept, we conducted a functional test using a lead-sensitive biosensor strain from our last year project. This experiment aimed to evaluate the effectiveness of alginate bead encapsulation in enhancing colorimetric detection performance.

The Testing of Alginate Beads with Lead-Sensitive Bacteria served as the core proof‑of‑concept procedure. In this test, the PbrR–pPbr lead biosensor (BBa_K5152004) was cultured and encapsulated within calcium alginate beads produced by combining sodium alginate and calcium chloride solutions. Beads containing the encapsulated biosensor were exposed to Lead (II) Nitrate solutions at concentrations ranging from 0 mM to 10 mM, then incubated at 37 °C for 24 hours. As shown in the figure below, the beads exhibited a gradual intensification of color with increasing lead concentrations.

Proof of concept

Bacterial beads after incubated in different concentrations of lead (II) nitrate LB Amp solution, after incubation at 37 °C for 24 hours. The beads were then extracted and put in distilled water for comparison. (left to right: 0mM, 0.01mM, 0.1mM, 1mM and 10mM lead (II) nitrate)

This observation demonstrates that encapsulated bacteria can still produce distinct, visible color changes, confirming that the alginate matrix effectively supports biosensor functionality and utility. The strong colorimetric response indicates that the bead system maintains bacterial viability and enables efficient signal transduction even within a confined environment.

These results verify the feasibility, practicality, and scalability of the Remedix encapsulation approach, establishing a strong proof-of-concept for future application to our tetracycline‑ and salicylate‑responsive E. coli strains once cloning and characterization are complete.

Tetracycline biosensor alginate beads were tested at 0, 100, and 200 ng/mL tetracycline (left to right).

In late September, near the end of the engineering cycle, we successfully cloned and validated BBa_25A81M7M, which features TetR expressed from a medium‑strength promoter (BBa_J23111) and medium RBS (BBa_B0032), with the pTet promoter regulating amilCP expression. As expected, cells developed a blue coloration in the presence of tetracycline, confirming the functional performance of the genetic circuit.

We then encapsulated the genetically modified bacteria in alginate beads and conducted preliminary functional tests. Although time limitations prevented a comprehensive analysis, these early results indicate successful integration of the biosensing construct within the Remedix platform, demonstrating its potential for field‑deployable biological sensing.

From the results, beads turned pale blue at 100 ng/mL and deep blue at 200 ng/mL tetracycline, while the control remained unchanged, confirming dose‑dependent amilCP expression and functional responsiveness of the biosensor system.

Conclusion

These experiments:

• Demonstrates utility and functionality
  demonstrating that alginate bead encapsulation does not hinder biosensor performance, the bacteria remain vialble and sensitive.
• Helps address similar challenges within the iGEM and wider synthetic biology community
  provixfing a general, easily adaptable platform for teams developing whole‑cell biosensors.

4. Development of the fluidic Modules

User Testing and User Feedback

Due to Hong Kong’s biosafety regulations, our team was unable to bring genetically modified (GM) E. coli outside the laboratory. To ensure our hardware remained practical and user‑oriented, we compensated by conducting simulated testing, expert consultation, and user demonstrations, presenting data and prototypes to professionals across sectors. These engagements provided valuable insights that directly guided hardware refinement and validation.

For hands-on user testing, we invited fish and pig farmers (our key industry stakeholders) to manually form alginate beads using non‑GM materials.

• Farmers found that bead making by hand was technically challenging and inconsistent.
• This feedback led us to develop a millifluidic bead‑formation device that automated the process, producing uniform, reproducible, and easy‑to‑use biosensor beads.
• Subsequent prototype demonstrations earned positive feedback from both users and academic experts, confirming improved usability and consistency.

Our team’s engagement with local fish and pig farmers

In response, we adopted a fluidic approach to automate bead generation. This method provided precise control, uniformity, and reproducibility, directly addressing farmers’ concerns about the slowness and repetitiveness of manual production. This evolution—from a biosensing concept to a functional, field-ready system—was driven by the need for greater reproducibility, user-friendliness, and automation, as identified by both academic experts and industry stakeholders. An in depth elaboration of the development of our millifluidic modules’ designs will be presented in the upcoming section.

Our fluidic device is composed of two modules: A mixing module and a bead‑formation module.

The Mixing Module: The main challenge was to mix two liquids of different viscosities, with sodium alginate being very viscous and the sample solution. Manual mixing often introduces air bubbles, which lead to floating and irregular beads. In contrast, our fluidic mixer uses channel geometry designs that enhance fluid contact time and surface area, achieving efficient mixing without bubble formation.

The Bead‑Formation Module: After thorough mixing, the liquid passes into the bead‑formation module, which precisely controls droplet size through flow rate regulation and nozzle design. This ensures uniform bead size and shape, which is crucial for maintaining consistent color intensity and reliable measurements.

Millifluidics prototypes were designed using the product development platform "onshape". Here is a link to the view only document of Remedix, with Onshape you have option to export .dxf file for laser cutting. Millifluidics Onshape Files

The design and testing of millifluidics by our team

The Mixing Module

We design our fluidic modules with reference to microfluidics and millifluidics studies. The study by Nishu and Samad (2023) presents an investigation of passive micromixers based on the split‑and‑recombination (SAR) flow principle using vortex‑generating hexagonal mixing units. Four micromixer configurations were modeled and compared by evaluating their mixing index, pressure drop, and energy cost over a Reynolds number range of 0.1–100.

Among the designs, Model 3—featuring dislocated connecting channels, demonstrate the highest mixing efficiency. The study concludes that SAR‑based geometries achieve excellent mixing performance with simple designs suitable for chemical and biological applications.

Reference:

Nishu, I. Z., & Samad, M. F. (2023). Modeling and simulation of a split and recombination‑based passive micromixer with vortex‑generating mixing units. Heliyon, 9 (4), e14745. https://doi.org/10.1016/j.heliyon.2023.e14745 ↩

With reference to this, we designed prototype 1 of the mixing module.

Prototype 1 of the Mixing Module

After reviewing additional literature, we noted that many passive micromixer designs, especially those used in lipid nanoparticle (LNP) synthesis, incorporate an “S”-shaped or serpentine channel geometry to improve mixing under laminar flow. According to Oliveres (2025), such curved pathways enhance mixing efficiency by generating secondary flows that promote fluid interaction without external energy input. Based on these insights, we developed a modified S‑shaped prototype to test its effectiveness in mixing high‑ and low‑viscosity liquids like sodium alginate and aqueous samples.

Reference:

Oliveres, R. (2025). Fundamentals of microfluidic mixing for LNP synthesis. Inside Therapeutics. Retrieved October 2, 2025, from https://insidetx.com/resources/fundamentals-of-microfluidic-mixing-for-lnp-synthesis

Based on these observations, we developed Prototype 2 and Prototype 3 of the mixing module, both derived from the initial “S”-shaped design but differing in channel length. We hypothesized that longer pathways might increase flow resistance, potentially preventing simultaneous entry of both liquids.

Prototype 2 of the Mixing Module

Prototype 3 of the Mixing Module

After fabricating the three prototypes using laser‑cut 3 mm acrylic sheets, each design was sandwiched between two flat acrylic plates to form enclosed millifluidic channels. For testing, 1:60 sodium alginate solution (red) and a blue dye aqueous solution were injected using a peristaltic pump at controlled flow rates. This procedure allowed us to visually observe flow behavior and assess how effectively each prototype design enabled simultaneous entry and mixing of both liquids within the microchannels.

(Coloured solutions were used for increased clarity, which do not affect mixing results. The corresponding solutions are colourless in practice.)

Proof of concept:

Prototype	Description
Prototype 1	As shown above, the pathways and voxels proved too large and wide, allowing the two liquids to slide past each other with minimal interaction and no significant mixing.
Prototype 2	The redesigned pathways exhibited excessive resistance, preventing both liquids from entering simultaneously—one typically dominated the channel.
Prototype 3	We balanced resistance and mixing efficiency by refining channel dimensions and flow paths, achieving uniform blending without dominance or slippage.

Conclusion

Based on the test results, we selected Prototype 3 for further use. It was developed by refining the earlier channel dimensions and flow paths to balance mixing efficiency and flow resistance. This optimized “S”-shaped layout, created through laser‑cut acrylic layers, enabled smooth flow and uniform mixing of the injected sodium alginate and dye solutions.

The Bead-Formation Module

After further review of relevant literature, we found that the emulsion method is widely used for droplet generation through flow agitation (Trantidou et al. 2018). Based on this approach, we selected the A (iii) flow‑focusing configuration from the study to design our own fluidic device, aiming to achieve consistent and uniform droplet formation.

Reference:

Trantidou, T., Friddin, M. S., Salehi‑Reyhani, A., Ces, O., & Elani, Y. (2018). Droplet microfluidics for the construction of compartmentalised model membranes. Lab on a Chip, Royal Society of Chemistry. https://doi.org/10.1039/c8lc00028j

With reference to their research, we came up with prototype 1 of the bead-formation module.

Prototype 1 of The Bead-Formation Module

On the other hand, we found that researchers in 2011 had developed an automated assembly for polymer bead production using a peristaltic pump system, which enabled controlled and continuous dripping of polymer solutions into a gelling bath (Atara et al. 2011). Although their setup operated at a macro‑scale and lacked optimization for small‑scale applications, it inspired us to miniaturize the dripping of alginate solution into calcium chloride solution through a millifluidic approach. Based on this concept, we designed spaced compartments across each board, interconnected by a pathway to collect the formed beads.

Reference:

Atara, S. A., Prajapati, B. G., Patel, V. P., & Desai, T. R. (2011). Design of novel assembly for automated production of polymer beads. Journal of Pharmacy Research, 4(7), 2254–2255.

Accordingly, we developed Prototype 2 and Prototype 3, with the main difference being the width of the channel. This modification was made because the flow rate of the calcium chloride solution directly influences the droplet formation process and, consequently, bead uniformity.

Prototype 2 of The Bead-Formation Module

Prototype 3 of The Bead-Formation Module

(A coloured solution was used for increased clarity, which does not affect bead formation. The corresponding solution is colourless in practice.)

Proof of concept:

Prototype	Description
Prototype 1	After further examining the paper in detail, we realised that we had misinterpreted the setup requirements. The configuration developed by (Trantidou et al. 2018) relies on two immiscible liquid phases to enable stable droplet generation. In our case, the Alginate-E. coli mixture came into direct contact with the calcium chloride solution, which immediately triggered crosslinking and formed a gel membrane at the interface. This premature solidification impeded proper emulsification, resulting in elongated, non‑spherical beads instead of spherical droplets.
Prototype 2	The pathway of prototype 2 was too narrow, hindering fluid flow and leading to inconsistent droplet formation and non‑uniform alginate beads.
Prototype 3	Prototype 3 with widened channels enhances solution flow and shortening non‑essential segments to improve system efficiency. As a result, the redesigned prototypes achieved uniform and consistent bead formation with greater encapsulation reliability, ensuring smoother operation and improved overall performance.

Conclusion:

With reference to the results, Prototype 3 was chosen for subsequent applications. It was developed by refining channel width and pathway geometry to achieve an optimal balance between flow resistance and mixing efficiency. It facilitated steady fluid movement and uniform mixing of the injected sodium alginate and calcium chloride solutions, ensuring consistent and reliable bead formation throughout operation.

Automated Fluid Transfer Using Mini Peristaltic Pumps

To enable automated and consistent fluid transfer within Remedix, we integrated mini peristaltic pumps to deliver solutions between millifluidic modules. Their gentle squeezing action protects delicate samples such as E. coli cultures while maintaining precise, stable flow rates for reproducible experiments. Since fluids only contact the tubing, cross‑contamination and cleaning requirements are minimized.

Compact and programmable, these pumps integrate seamlessly with other automated components like OD monitoring and temperature control. Their easy‑to‑replace tubing supports sterile operations and prevents biofilm buildup. Together, these features make peristaltic pumps ideal for ensuring accuracy, sterility, and reliability in automated biosensing workflows.

The peristaltic pump and respective boards for controlling its action

The components of the peristaltic pump

5. The Evolution of Remedix: From fluidic Device to Integrated Platform

Our journey toward developing Remedix continues with extensive consultations with professionals across multiple disciplines. We gathered valuable feedback from policy and academic experts, including Mr. Mok Wing‑cheong (Director of DSD), Professor Benjamin J. Cowling, and Professor Frank Lam, whose expertise played a crucial role in shaping the direction of our hardware. These experts emphasized the need to transform our project from a single fluidic bead‑generation device into a portable, integrated laboratory platform capable of culturing bacteria, forming beads, and performing on‑site detection. Their recommendations guided us to expand Remedix’s scope beyond agriculture, allowing for its application in wastewater monitoring, pollution tracing, and healthcare diagnostics.

Our team’s engagement with professionals of various departments

Drawing inspiration from our team 2024’s “Metalytic” machine, we re‑envisioned our original fluidic device concept into a complete, automated, and consistent biosensing platform designed for both bead generation and colorimetric detection.

Remedix integrates three main components into one seamless system:

• A stirring incubator for culture of E. coli.
• An automated bead‑formation module based on fluidic technology for consistent encapsulation.
• A smart detection unit capable of both online and offline operation, equipped with an LCD interface for real‑time display and SMS/WhatsApp notification for instant alerts.

6. E. coli Culture Incubator System

We integrated the GM E. coli culturing process into Remedix to create a complete, user‑friendly biosensing platform. Building on our team’s 2024 portable bacterial culture system “Metalytic,” which successfully demonstrated the cultivation of E. coli, we incorporated valuable insights from last year’s team members and judges’ feedback to enhance our system design. Specifically, we introduced a calibrated optical density (OD) monitoring system this year to accurately track bacterial growth and determine the optimal point for downstream processing. During the development of our culture system for reporter‑expressing GM E. coli, we consulted Prof. Tam Fung‑yee and Prof. Tsui Tsz‑ki, who provided valuable suggestions that helped refine our culturing workflow and improve reliability.

How It Works

To ensure safety and containment, the system includes a UVC sterilization module that disinfects the culture chamber before each experiment. This step eliminates any remaining bacteria, preventing contamination—especially important when handling GM E. coli.

Bacterial growth is monitored using optical density (OD), which measures how cloudy the culture becomes as E. coli multiplies. A LED light source and ESP32 camera detect how much light passes through the broth—less light indicates more bacterial growth. We utilized a regression model to build the calibration curve, establishing a reliable mathematical relationship between the measured signal and the target concentration. This setup enables real‑time tracking of growth stages and helps determine the best harvest time for maximum yield.

A magnetic stirrer continuously mixes and aerates the culture, ensuring even nutrient distribution and sufficient oxygen supply for optimal bacterial metabolism. This promotes accurate readings and consistent growth. Meanwhile, temperature control, managed by an ESP‑32E microcontroller, DHT11 sensor, heater, and fan, maintains ideal culture conditions. Stable aeration, mixing, and temperature regulation prevent measurement errors and support healthy bacterial development. We employed a PID controller algorithm, leveraging its theoretical foundation of proportional, integral, and derivative actions, to achieve stable temperature regulation.

7. Colour Sensor Using AI Vision

We use AI and machine learning to quantify chromoprotein expression in genetically modified E. coli as an indicator of pharmaceutical waste in water. The bacteria produce color‑changing chromoproteins in response to contaminants, and AI‑driven image analysis measures color intensity and patterns from bead images. Trained on datasets of known concentrations, the machine learning model correlates color variation with pollutant levels. This allows the system to quickly and accurately detect and measure pharmaceutical waste in water samples.

How It Works

To monitor color changes in our engineered E. coli biosensors after the encapsulated bacterial beads come in contact with the tested sample, we deployed ESP32-CAM modules as webcams. However, the sample matrix’s natural hue could interfere with the beads’ colorimetric signals. To address this, we designed a custom sieve assembly that lifts the beads above the liquid, ensuring clear visibility. Its perforated platform maintains a monolayer distribution, minimizing bead overlap for consistent imaging.

The cameras track real-time color shifts and upload images to a cloud-based AI server (hosted on PythonAnywhere). Our model processes these images, precisely quantifying color intensity to detect unsafe pharmaceutical waste levels. If thresholds are exceeded, the system triggers instant alerts via SMS/WhatsApp.

The ESP32-CAMs will transmit frames to the server repeatedly. This integration of hardware, AI, and IoT enables rapid, reliable contamination monitoring.

Camera (ESP32 Cam Module)

Proof of concept

The images below demonstrate the AI training process for color recognition of alginate beads. During testing, we used alginate beads infused with different color dyes to create a diverse training dataset. Each bead in the dataset was manually annotated and categorized according to its visible chromoprotein color (e.g., brown or purple).

Training of the AI model, where colour dye alginate beads were annotated to teach the system accurate color recognition and classification

The labeled images were used to train the AI model to automatically detect and classify alginate bead colors in real time. The bounding boxes indicate the regions recognized by the training software, showing how the system learns to distinguish subtle variations in hue and intensity. As demonstrated in the training results, the AI model successfully differentiates between brown and purple alginate beads using annotated datasets. Although these colors appear visually similar to the naked eye, the model accurately identifies them across varying lighting conditions and image backgrounds. This robust workflow enables the AI vision module to precisely quantify color changes, providing a strong foundation for automated and objective biosensor analysis.

Testing of the AI model with brown and purple colour dye beads.

Training results showed that, despite the colors appearing visually similar, the AI system accurately distinguished them across different lighting and backgrounds, proving its precision in color‑based biosensor analysis.

Alginate beads containing the tetracycline-sensitive biosensor were incubated in 0, 100, and 200 ng/mL tetracycline at 37°C for color development. Despite faint blue coloration, the AI model accurately distinguished samples.

Towards the end of the engineering cycle, after we successfully cloned and verified the function of the tetracycline biosensor, the alginate beads were tested at 0, 100, and 200 ng/mL tetracycline. We then applied AI-based image analysis to distinguish between the color responses. Although the blue coloration developed by the reporter was not particularly intense, the AI model was still able to accurately differentiate the samples based on subtle variations in color intensity, confirming the system’s responsiveness and the viability of AI-assisted detection.

Details of AI Model Development and Training

We developed our custom object detection model using the YOLOv5 framework for chromoprotein color detection in alginate beads.

Dataset Preparation

• Annotated images using makesense.ai web-based tool
• Applied bounding boxes around chromoprotein-colored beads
• Exported annotations in YOLO format (normalized coordinates)
• Organized into train/ and val/ directories

Training Environment

• Utilized Google Colab with Tesla T4 GPU
• Cloned YOLOv5 repository and installed dependencies
• Created dataset.yaml configuration file

Training Process

• Initialized with pretrained weights
• Training parameters: image resize to 640×640, batch size 16, epochs 100
• Real-time monitoring via TensorBoard (precision, recall, mAP metrics)
• Automatic best model checkpoint saving

Deployment

• Final best.pt model deployed to PythonAnywhere cloud server
• Integrated with ESP32-CAM for real-time bead color detection
• Iterative improvement through edge case identification and dataset expansion
• Advanced users can download datasets from our platform to train and deploy their own custom models.

For detailed methodology, we followed: https://docs.ultralytics.com/yolov5/tutorials/train_custom_data/

8. Real‑Time SMS Notifications and Instant Alerts

Remedix utilizes IoT technology to provide immediate alerts when pharmaceutical waste contamination exceeds safe thresholds. Once detected, the system automatically sends SMS or WhatsApp notifications to users, ensuring prompt awareness and response. Each alert message also includes contact information for relevant authorities, such as Hong Kong’s Centre for Food Safety and the Water Supplies Department, to facilitate quick reporting and coordinated follow‑up actions.

How it works

We utilize the Twilio API to integrate programmable messaging into our application. We have written the API calling functions to our code. When the AI detects a color change in the beads, it will send a notification to users via the Twilio API, warning them that the tetracycline or salicylate concentration has exceeded safe thresholds.

Proof of concept

Notifications on SMS/ WhatsApp

LCD touch Screen (ESP32-C6-Touch-LCD-1.47)

For immediate on-site visibility, the system features an integrated LCD touch screen that displays real-time detection results. If thresholds are breached, a clear warning appears, ensuring awareness even without mobile devices.

The ESP32-C6-Touch-LCD-1.47’s functionality was confirmed through testing. The corresponding code for the ESP32-C6-Touch-LCD-1.47 that controls the aforementioned actions is documented in the HongKong-JSS GitLab repository.

9. Website and Software for Monitoring Backend Processes

We have developed a comprehensive multifunctional website and mobile app that enables seamless system operation, real-time monitoring, and rapid response to contamination events. The functional management website is hosted on PythonAnywhere, an online integrated development environment and web hosting service built on the Python programming language.

Web Platform Operations:

• AI model training for continuous detection improvement
• Live streaming of images captured by ESP32-CAM modules
• Adjusting color detection standards to calibrate biosensor thresholds
• Configuring SMS notification intervals for alert frequency control
• Generating predictions for detection results using trained models
• Accessing detailed system logs for troubleshooting and maintenance

The functional management website

Meanwhile, we developed a user-friendly app using figma with features:

The management app with user-friendly graphics

The Mobile App Features:

Real-Time Status Monitoring

• Live, color-coded dashboard displaying current contaminant levels and overall system health
• WebSocket connection maintains persistent link between app and cloud server for instant updates
• Automatic display refresh when YOLO-based AI completes image analysis—no manual refresh needed

Instant Contamination Alerts

• Dual-channel alerts: push notifications (Firebase Cloud Messaging) and SMS (Twilio API)
• Server-side rules engine continuously monitors detection data
• Simultaneous triggering of both notification channels when thresholds are breached
• One-tap incident reporting to authorities for immediate response

Predictive Contamination Forecasting

• AI-generated charts predicting contaminant concentration trends for next 48 hours
• LSTM machine learning microservice runs on cloud server
• Periodic retraining on historical data for improved accuracy
• Computation offloaded from mobile devices to powerful cloud infrastructure

Automated Compliance Reporting

• Single-tap generation of professional PDF reports for any selected time period
• Server-side API queries SQL database and formats data using headless PDF libraries
• Reports ready for regulatory submission
• Resource-intensive processing handled on server, not user devices

Centralized Multi-Site Management

• Interactive map interface displaying all remote sensor stations
• Color-coded pins indicating real-time status of each location
• GPS coordinates and latest analysis results stored in cloud database
• Geospatial API enables oversight of distributed monitoring networks from single interface

Diagram summarized features of our apps.

Together, the web platform and mobile app form a complete end-to-end solution—from AI-powered detection to real-time alerts, predictive analytics, compliance documentation, and multi-site coordination—making Remedix a practical, deployable system for environmental biosensing.

Programme Involved in Remedix

1. MicroPython (ESP32 Boards)
Controls camera (ESP32-CAM) for image capture and motor drivers (L298N) for pumps/bead automation.

2. Flask and Python (Backend/Website)
Hosts AI processing, live streaming, and adjustments on PythonAnywhere.

3. C++ (UI Components)
Manages LCD touch screen (ESP32-C6-Touch-LCD-1.47) for offline results and controls.
(ref: ESP-IDF (https://www.waveshare.com/wiki/ESP32-C6-Touch-LCD-1.47))

4. Figma (App)
Prototypes mobile interface for remote monitoring and notifications.
(App link: https://www.figma.com/make/b20x4Mr1VOyLKDHC9RAq5G/Remedix-AI-App-Design?fullscreen=1)

10. Affordable Innovation: Remedix Makes Advanced Water Biosensing Accessible

The total development cost of Remedix is approximately USD $50, making it an extremely affordable alternative to traditional laboratory instruments. In comparison, professional analytical equipment such as ICP‑MS costs between USD $50,000 and $250,000, restricting access to well‑funded institutions. Remedix significantly reduces the financial barrier to environmental biosensing, enabling routine, decentralized testing by farmers, technicians, and communities. This low‑cost, portable, and user‑friendly system bridges the gap between laboratory‑grade precision and everyday public accessibility, advancing real‑time detection of pharmaceutical and antibiotic contaminants in water.

All prototype was designed using the product development platform "onshape". Here is a link to the view only document of Remedix, with Onshape you have option to export .dxf file for laser cutting. Prototypes Onshape files

Component	Spec	Function	Price
Temperature Control System
DHT 11 Sensor	Temperature: ±2℃ Humidity: ±5％RH DC 3~5.5V	Continuous temperature sensing and data transmission using the built-in MCU to the main board via the signal pin.	$ 0.15
Cooling Fan	Size: 10cm*10cm Voltage: 12V Max RPM: 3000	Ventilation and cooling when high RPM.	$ 0.43
Ceramic Heater	Size: 50*20*5mm 5V/50℃ (2~6W)	Heating element	$ 0.43
Optical Density (OD) Monitoring System
Camera (ESP32 Cam Module)	Wi-Fi: 802.11b/g/n/e/i Cam: OV2640 Voltage: 5V	A webcam that captures images and uploads them to the management server	$ 2
LED Light Bulb	Voltage: 3.3V Brightness:100 lm/W	Light source for OD	$ 0.7
Stirring Unit
R300C DC motor	Voltage: 1.5~6 V Speed: 3500 ~ 7000 rpm	Motor for stirring	$ 0.4
Magnetic rod	Size: 0.6* 2cm	Rod for stirring	$ 0.11
Sterilizing System
Low-power UVC light	Power: 3w Size: 32mm Voltage: 5V	Sterilization before culturing	$ 1.47
Peristaltic Pumps Unit
Peristaltic Pump	Voltage: 12V Rate: ≥70ml/min	Transfer an accurate amount of liquid	$ 17.76 (6 * $ 2.96)
Silicone Tube	Inner radius: 3mm Outer radius: 5mm	Transfer pathways	$ 0.2
2* L298n & ESP32EA-WROOM‑32E (source: Hello STEM platform)	motor drivers & controller board	Pump system control and Main board controller	$ 3 (2 * $ 0.75 + $ 2)
Detection System
Camera (ESP32 Cam Module)	Wi-Fi: 802.11b/g/n/e/i Cam: OV2640 Voltage: 5V	A webcam that captures images and uploads them to the management server	$ 2
ESP32-C6-Touch-LCD-1.47	2.4GHz Wi-Fi BLE 5 LVGL GUI	An interactive screen that allows users to input parameters to the machine and read the result of the machine	$ 9.6
Sieve	radius: 3.5cm	Level up and hold the beads above the liquid level to make it easier to capture images.	$ 0.2
Servo Motor	FS90R turning angle: 360 deg	Connect to a rope that can be used to level the sieve up and down.	$ 1.7
Others
Fiberboard	/	Laser cut them to build the framework of the whole machine.	$ 2.98 (3 * $ 0.993)
Wires	/	Connections	$ 1
Power Bank	5V, type-c connetion	Serve as power supply to the whole system	$ 5
/	/	/	Total: $ 49.13

Evaluation of Remedix Hardware Through Stakeholder Feedback

We presented the final Remedix design to a wide range of stakeholders, including medical professionals, university professors, and local farmers. These consultations provided valuable real‑world insights into usability, safety, and operational requirements. Based on their feedback, we developed a clear understanding of both the strengths and limitations of the Remedix hardware system.

Strengths

1. Low‑Cost and Accessible

• The complete hardware setup costs only around USD $50, making it significantly more affordable than traditional laboratory instruments like ICP‑MS (USD $50,000–$250,000).
• Its low cost promotes widespread adoption by farmers, small laboratories, and environmental organizations for routine water quality monitoring.

2. Integrated, Portable Design

• The system integrates culture, encapsulation, and detection functions into a single compact platform.
• Designed for mobility and field usability, it can be operated without specialized laboratory facilities.

3. Automation and Ease of Use

• Incorporates millifluidic modules for automated bead production, reducing manual labor and potential human error.
• The peristaltic pump system ensures consistent and precise liquid handling across modules.

4. Expanding Potential of Remedix Through Modular Design

• Proven core functions; with minor modifications, Remedix can transform into a low‑cost shaking incubator, AI plate reader, or fluorescence signal reader.
• Modular design supports future module integration, promoting affordable and accessible synthetic biology.

5. AI and IoT Integration

• Uses ESP32‑CAM modules and AI‑driven image recognition to accurately detect and quantify color changes.
• IoT connectivity allows real‑time notifications via SMS or WhatsApp, enabling continuous remote monitoring.

Limitations

1. Fabrication and Durability Issues

• The laser‑cut acrylic, three‑layer construction may allow small fluid leaks and compromise durability.
• The prototype has not yet been optimized for long‑term field deployment or harsh environmental conditions.

2. Limited Environmental Adaptability

• Sensors and optical components are calibrated for stable indoor conditions, and performance may vary with temperature, humidity, or light changes.

3. Single‑Sample Throughput

• The system currently supports only one test at a time, limiting scalability for large‑scale or multi‑site environmental assessments.

4. Testing Constraints

• We have not yet tested the system with the salicylate‑sensitive bacterial strains due to time constraints.
• Although we cloned and confirmed the expression and function of the TetX tetracycline‑degrading strain, it has not yet been tested in the alginate beads.
• Nevertheless, a proof‑of‑concept has been successfully demonstrated, validating the underlying design and detection mechanism.

5. Sensor and Optical Calibration Challenges

• The AI camera remains sensitive to environmental lighting, bead positioning, and background interference.
• Regular calibration is required to maintain quantitative accuracy and consistency over prolonged use.

Future Directions

Building on the lessons learned from the current prototype and the limitations identified, future work on Remedix will focus on improving its reliability, adaptability, and validation through comprehensive testing. The team also plans to expand the system’s design framework into new applications within synthetic biology.

1. Experimental Validation with our GM E. coli Strains to Confirm Detection and Degradation Capability

To establish the full biological and analytical capabilities of Remedix, future experiments will focus on validating the system using engineered bacterial strains designed to detect and degrade tetracycline.

• Conduct more experiments using tetracycline- and salicylate-sensitive strains to evaluate the system’s biosensing performance.
• Compare the hardware’s detection results with standard chemical assay methods for quantitative verification.
• Assess the response time, sensitivity, and operational stability of both the bacteria and detection hardware to ensure consistent and reproducible results.
• Future work will test the tetracycline-degrading strains after adding a secretion tag to enable extracellular activity. Prior studies show enzymes can diffuse from alginate-encapsulated E. coli, supporting this approach’s feasibility.

2. Environmental Adaptability Enhancement to Ensure Reliable Performance in Outdoor Settings

To ensure robust performance under diverse environmental conditions, future versions of Remedix will be designed and calibrated for outdoor field applications.

• Calibrate optical sensors and image detection systems to maintain accuracy under varying temperature, humidity, and ambient lighting conditions.
• Integrate automatic environmental compensation algorithms into the AI model to stabilize output signals and maintain consistent readings during field use.
• Encase the device in weatherproof or dust‑resistant housing to protect sensitive electronic components and optics from environmental degradation.
• Conduct extensive field trials across rural and industrial sites to assess long‑term reliability, user handling, and data stability under real‑world conditions.

3. Expansion into Broader Synthetic Biology Applications

Beyond tetracycline detection, the foundational design of Remedix presents a versatile platform adaptable for a variety of synthetic biology applications.

• Integrate different engineered microbial strains with specific sensing and degradative abilities to address varied environmental challenges.
• Collaborate with academic and industrial partners to expand Remedix into a modular biosensing ecosystem, enabling community‑level environmental monitoring and educational use in synthetic biology outreach.