I am a Data Scientist Manager and AI Engineer passionate about building real-world AI-powered solutions. My expertise spans Generative AI, Agentic Systems, Computer Vision, Industrial IoT, and Robotics, where I combine research-backed approaches with practical, production-grade deployment.

I have designed and delivered projects ranging from autonomous AI firmware for microcontrollers and multilingual conversational agents to real-time analytics dashboards and AI-driven mobile applications. My focus is always on building scalable systems that bridge the gap between software intelligence and physical hardware to create real impact.

I actively work with:

• Agentic AI & LLMs: CrewAI, LangGraph, Google Gemini, Groq, Llama 3
• Edge & Hardware: ESP32, FreeRTOS, PlatformIO, TFLite
• Cloud & MLOps: AWS, GCP, MLFlow, Docker
• Full-Stack AI: FastAPI, Flask, Streamlit, Flutter

My strength lies in end-to-end AI development, from building robust ML pipelines to developing user-centric interfaces and embedded hardware solutions, always driven by continuous learning and a commitment to excellence.

📧 Let's Connect: abhimanyus1997@gmail.com

• MicroClaw - Agentic Firmware for ESP32 (2026)
  Tech Stack: C++, PlatformIO, FreeRTOS, LittleFS, Gemini/Groq API, Python, FastAPI
  Firmware that turns an ESP32 into an LLM agent. Gemini or Groq drives tool calls for GPIO control and scripting via FreeRTOS tasks, WiFi/BLE scanning, and persistent LittleFS memory. Chat over an on-device web server, Telegram, or serial; a FastAPI desktop manager handles flashing, configuration, and monitoring.
  View Docs · Repository

• AdiVox - Speaker-Aware Transcription (2026)
  Tech Stack: FastAPI, Gemini 2.5 Flash (google-genai), React, TypeScript, Vite, SQLite
  Uploads audio to a FastAPI backend that uses Gemini 2.5 Flash for speaker diarization with timestamps. A React frontend presents the transcript, and sessions are saved to a local SQLite library.
  Repository

• AWS Bedrock Bridge for Copilot Chat (2026)
  Tech Stack: TypeScript, VS Code Extension API, AWS SDK v3 (Bedrock Converse API)
  VS Code extension that registers Amazon Bedrock models (Claude, Llama, Nova, DeepSeek) as a GitHub Copilot Chat provider, with streaming, tool calling, cross-region inference profiles, custom model ARNs, and an account model-access check.
  Repository

• HealthBuddy - AI Wellness Companion (2026)
  Tech Stack: Flutter, Dart, Health Connect / Apple Health, Groq API (GPT-OSS, Llama 3.3/4), fl_chart
  Cross-platform health app that syncs steps, sleep, heart rate, calories, and weight from Health Connect or Apple Health. NutriGPT, a Groq-powered chat with selectable models, generates meal plans and wellness advice, with charts, an activity heatmap, and PDF export.
  Repository

• Jivika - Career Workspace for Job Seekers (2026)
  Tech Stack: FastAPI, React 19, TypeScript, Vite, Tailwind CSS, Recharts, Gemini / OpenAI / Groq / OpenRouter
  Bring-your-own-key career platform: resume-to-job matching that blends LLM evaluation with deterministic keyword coverage, ATS-style CV review, LinkedIn optimization, career paths with estimated salaries, interview prep with answer feedback, cover letters, and an application tracker. Keys and user data stay in the browser; the stateless API validates every model response against typed schemas.
  Repository

• LangChain One (2023)
  Tech Stack: LangChain, CTransformers (GGUF), Ollama, FAISS/Chroma, Streamlit, Chainlit
  LLM application suite running local models (OpenHermes-2.5 Mistral 7B) with a Chainlit chat, a PDF question-answering DocBot backed by vector stores, and a Diffusers text-to-image page.
  Repository

• Article: "The Evolution of Object Detection, YOLO & NAS Revolutionizing Computer Vision"
  Published on IndiaAI.gov.in
• Paper Presentation: "Subaqueous Crack Detection using Fusion-Based Image Enhancement for Dam Structure"
  Presented at the 4th International Conference on River Corridor Research and Management (ICCRM) 2024, hosted by IIT Guwahati & IIT Jammu

Let's connect to build innovative solutions together!