AI Builder • Backend Developer • Curious by Default
I build AI tools, backend systems, and automation that people can use.
From crop-market workflows to AI chess benchmarks — I learn by building, shipping, and refining.
An AI-powered bot that turns raw WhatsApp crop offers into structured, translated updates for buyers. Deployed locally and used by 1,00+ agricultural brokers across Telangana.
✨ Key Features
- Extracts crop-offer details from incoming messages using Gemini.
- Translates offers into target languages, including English and Telugu.
- Routes updates to buyer groups by crop type and language.
- Queues message processing, maintains logs, and reports processing failures.
- Provides a local dashboard to monitor activity and toggle automation.
Tech Stack: JavaScript, Node.js, Express, Gemini API, whatsapp-web.js, Socket.IO
A platform for studying how language models make sequential decisions through competitive chess. Compares LLMs with Stockfish to investigate decision quality, rule compliance, and consistency.
** Key Features**
- Matches AI players from multiple providers against each other or Stockfish.
- Uses modular adapters for Gemini, OpenAI, and Claude.
- Streams game state to the frontend through WebSockets.
- Validates moves and handles invalid responses with retries and fallback logic.
- Separates game orchestration, player adapters, and provider integrations.
Tech Stack: Python, FastAPI, WebSockets, python-chess, Stockfish, SvelteKit, TypeScript
The frontend and additional benchmark metrics are in development.
Explore the basic chess demo →
An experimental system exploring how AI agents can coordinate software development. A shared orchestrator brings management, development, and QA agents together around a user task.
** Current Structure**
- Separate agents for management, development, and quality assurance.
- An orchestration layer to coordinate the workflow.
- LLM integration and file tools to support software tasks.
Tech Stack: Python, LLM APIs, Agent Orchestration
Early-stage exploration.
| Project | What I explored |
|---|---|
| 🤟 Sign Ease | Hand-gesture image classification with TensorFlow and a FastAPI prediction endpoint. |
| 🪽 Watchful Wings | A Python safety analytics prototype exploring threat detection and real-time alerts. |
| 💳 Mahatma Coin Wallet | A React and TypeScript wallet prototype exploring Aadhaar-based authentication. |
- AI evaluation: how language models behave when decisions must follow explicit rules.
- Practical automation: turning unstructured messages into useful workflows.
- Agent systems: coordinating planning, implementation, and review.
Have an idea for an AI tool, a backend system, or a workflow worth automating?
avinashsana2@gmail.com •
GitHub
