AI Engineer building GenAI products, agents, automation systems, and document intelligence workflows
- 🤖 AI Engineer based in Bengaluru, India, focused on GenAI, LLM applications, RAG, agents, and automation.
- 🧠 I enjoy building practical AI systems: offline AI assistants, document extraction, web crawling/search, YouTube automation, and autonomous workflows.
- 📱 Creator of Anvit: Local Agentic RAG, a private on-device AI document assistant for PDF and Word files.
- 🛠️ I work across Python, FastAPI, TypeScript, React, Android, Docker, vector databases, embeddings, and LLM orchestration.
- 🚀 Currently exploring production-ready AI agents, GenAI SaaS products, and privacy-first local AI apps.
- 📫 Reach me at: likhithv02@gmail.com
| Project | What it does | Stack / Focus |
|---|---|---|
| Anvit: Local Agentic RAG | Private offline Android AI assistant to chat with PDF and Word documents. Uses on-device Gemma via Google LiteRT, hybrid retrieval, agentic RAG, corrective RAG, and cited sources without sending documents to the cloud. | Android, local LLMs, Gemma, LiteRT, Agentic RAG, BM25, embeddings, privacy-first AI |
| DocExtract | Modern document extraction app for government IDs and invoices using LlamaParse AI with a FastAPI backend. | TypeScript, React, FastAPI, MongoDB, Docker, parsing |
| WebNexus | Self-contained web crawling and document search system with AI-powered context retrieval. | Python, RAG, embeddings, FAISS, scraping, MCP |
| Youtube-automation | Automates YouTube video creation from an idea: storyline, images, narration, and MoviePy video generation. | Python, agents, automation, AI video, MoviePy |
| ShowShepherd | Autonomous media acquisition and streaming assistant that understands natural-language media requests. | Python, Docker, automation, streaming, Plex |
| Groq-Function-Calling | LLM function-calling experiments integrating Groq with tools such as Gmail workflows. | Python, Groq, LLM agents, function calling |
| DocQA / DocQA-new | Document question-answering and form understanding experiments with classification and retrieval. | Docker, Donut, FAISS, Groq, fine-tuning |
AI / ML / GenAI: LLMs, RAG, Agentic RAG, Corrective RAG, embeddings, vector search, hybrid retrieval, agents, function calling, document AI, prompt engineering, LangChain, LlamaIndex, Transformers, Groq, FAISS
Backend / Infra: FastAPI, REST APIs, Docker, Docker Compose, MongoDB, PostgreSQL, Linux, GitHub Actions
Mobile / Frontend: Android, Kotlin, React, TypeScript, JavaScript, HTML, CSS
I don't measure progress only by commit volume. A lot of my work happens in research, experimentation, prototyping, model/tool evaluation, and architecture decisions before code lands in a repository.
- 📚 Researching AI systems, agent workflows, local LLMs, RAG patterns, and GenAI product ideas.
- 🧪 Experimenting with prompts, retrieval pipelines, model behavior, and automation workflows.
- 🏗️ Once ideas are validated and tested, I use AI coding tools like Claude Code and Codex heavily to move fast from prototype to product.
- 🔐 Strong interest in privacy-first AI, especially local/offline assistants like Anvit.
I'm always interested in AI engineering, GenAI SaaS ideas, automation, and applied ML projects.
