TanakPod

TanakPod, the AI-powered smart agent revolutionising tropical farming through real-time soil, crop, and climate intelligence for higher yields and sustainable plantations.

Comment

Inspiration

In Malaysia, especially in Sabah, many small and medium-sized farmers still rely on manual observation and experience to make key farming decisions such as irrigation, fertilisation, and harvest timing. Unlike large plantations, these farmers rarely have access to modern agri-tech tools that can help them monitor soil health, weather, and crop readiness in real time.

We wanted to change that. Our inspiration for TanakPod came from seeing how data-driven agriculture could drastically improve productivity, resource efficiency, and sustainability and yet remains inaccessible to those who need it most.

By leveraging AI, IoT sensors, and automation through AWS, our goal is to empower these farmers with simple, affordable, and intelligent tools that act like a “farm assistant,” guiding decisions from soil to sky.

Just a disclaimer that this is still in our ideation phase and more research is required to make this idea a reality. For now we use data that are available online and applied the calculation in our lambdas.

What it does

TanakPod is an AI-driven precision-farming assistant that helps oil palm farmers make smarter decisions on the field.

Core features:

  1. Collects sensor data - Each pod node captures soil moisture, electrical conductivity, nutrient index, and environmental data from the field.
  2. Integrates external weather data - Automatically pulls live forecasts (via the Open-Meteo API) to include rainfall predictions.
  3. AI decision engine - Data is sent to an AWS Lambda “Ingest” function, which then forwards it to the “Decision” Lambda powered by Claude 3 Haiku (AWS Bedrock). HARVEST if NDVI ≥ 0.72 and rain < 15 mm IRRIGATE if soil moisture < 0.18 APPLY FERTILISER if soil nutrients < 0.4 or EC < 1.2 Otherwise MONITOR
  4. Real-time alerts - The chosen action is logged in Firebase and sent instantly via WhatsApp (Twilio API).
  5. Interactive web dashboard – Users can enter data, preview 24-hour rainfall, and view AI decisions in real time. Our main idea is actually to use IoT such as drones and IoT but we don't have the IoT yet so we use a website as simulation in this hackathon. The website requires users to input the data and the TanakPod will give advices based on the input.

How we built it

Our architecture is serverless, modular, and built entirely on AWS.

Frontend A responsive web dashboard built with HTML, CSS, and JavaScript. Integrates directly with the ingest Lambda and displays rainfall data from the Open-Meteo API.

Backend and Cloud

  1. AWS Lambda
  2. ingest_fn: Handles input, fetches rainfall data, and invokes the decision layer asynchronously.
  3. decision: Processes analysis using Claude 3 Haiku, saves results to Firebase, and sends WhatsApp alerts.
  4. AWS Bedrock (Anthropic Claude 3 Haiku) for AI reasoning and structured JSON output.
  5. Firebase Realtime Database for decision logging.
  6. Twilio API for WhatsApp alerts.

Integrations

  1. Open-Meteo API for real-time rainfall forecasts.
  2. Environment variables securely store Firebase credentials and Twilio keys.

Challenges we ran into

Our main challenge was time. We started the project only ten days before the submission deadline, and both of us were completely new to AWS services. This meant we had to learn cloud fundamentals, IAM permissions, and Lambda integrations while actively building the project. We also faced difficulties in configuring permissions between AWS Lambda, Bedrock, and Firebase, and understanding asynchronous communication between functions. The Twilio sandbox presented another challenge with its strict 24-hour user message window, which initially prevented our WhatsApp alerts from being delivered. Additionally, integrating multiple APIs and ensuring seamless data flow without latency issues required careful debugging. Despite these obstacles and our limited timeline, we managed to deliver a fully functional, end-to-end prototype that runs reliably in real time.

Accomplishments that we're proud of

We’re incredibly proud that, within a short timeframe, we managed to build and deploy a fully functional AI-driven agri-tech platform that bridges cloud intelligence with real-world impact. As complete beginners to AWS, we successfully implemented a multi-function Lambda architecture connected to Bedrock, Firebase, and Twilio, achieving live decision-making and automated WhatsApp alerts. We also designed a futuristic yet simple user interface that allows manual input testing, rainfall previews, and immediate AI responses. Most importantly, we proved that accessible, low-cost precision-farming tools can be built using cloud technologies, potentially transforming how smallholder farmers make daily agricultural decisions.

What we learned

Through this project, we learned how to design a serverless, event-driven architecture using AWS Lambda and asynchronous invocations. We gained experience in prompt engineering, ensuring consistent JSON outputs from Claude 3 Haiku, and learned the importance of secure configuration through environment variables and IAM roles. Integrating multiple APIs such as Open-Meteo, Twilio, and Firebase has taught us how to manage cloud workflows efficiently while maintaining low latency. On the frontend side, we deepened our understanding of designing interactive dashboards for IoT applications. Beyond the technical knowledge, we also learned how to manage a complex project under tight deadlines, collaborate effectively, and rapidly iterate to turn an idea into a working AI-powered prototype.

What's next for TanakPod

TanakPod is a project that is deeply personal to us, and we are determined to see it evolve from concept to reality. What began as an idea to empower smallholder farmers through accessible AI and IoT technology has now gained real momentum as we have been selected to join an accelerator program that will guide us through the full journey from ideation to incubation, commercialisation, and long-term sustainability. While our initial focus is on oil palm plantations, we envision expanding into other agricultural sectors such as rubber, cocoa, and even aquaculture, creating a unified ecosystem of intelligent, data-driven farming.

Beyond improving productivity, we aim to ensure that farmers can adhere to the EU Deforestation Regulation (EUDR) by providing transparent, verifiable data trails for land use and crop sourcing. As we scale, we plan to integrate blockchain technology into TanakPod’s infrastructure to enable traceability and secure record-keeping across every stage of the agricultural supply chain. This will not only enhance compliance and accountability but also help farmers demonstrate sustainable practices and gain access to international markets.

Ultimately, TanakPod seeks to become more than just an agri-tech solution, it will be a social enterprise platform that uplifts local communities, strengthens sustainability compliance, and builds a new standard of trust and transparency in agriculture across Southeast Asia.

Built With

Share this project:

Updates