Generative AI • Large Language Models (LLMs) • RAG • AI Agents | Data Science • Machine Learning | Python • SQL | Cloud (GCP • AWS) • Docker | Open Source Contributor

• 🔭 I’m currently working on: Building Applied AI & Agentic workflows, and contributing to Open Source via GSSoC.
• 🌱 I’m currently learning: Advanced Java (transitioning to Spring/Spring Boot), MongoDB, Advanced RAG architectures, LangGraph, and Genkit.
• 💬 Ask me about: LLM frameworks, "Value-First" Data Science, tech community building (GDG, API Community Pune), or theatre scripts!
• 😄 Pronouns: Dr. Udayan
• ⚡ Fun fact: I blend tech and leadership — I build impactful AI inference systems while leading as the Google ELP Co-Head of Operations (Pune Region), AI Developer for E-Cell, and contributing to the NASA OSDR!

• 🌌 NASA OSDR Contributor: Utilizing AI/ML frameworks to analyze high-dimensional space biology data within NASA's Open Science Data Repository under the AIML AWG.
• ⚡ Tech Leadership: Appointed as the Google ELP Co-Head of Operations (Pune Region) and serving as the AI Developer for DYPCOEI E-Cell.
• 🛠️ Hackathons & Credentials: Participant in the Smart India Hackathon (SIH) and recently earned the Google Gemini Certified University Student credential.
• 🌐 Open Source Advocate: Active contributor to the GirlScript Summer of Code (GSSoC) 2026.

Languages:

Databases:

Generative AI, ML & Data Science:

Frontend Development:

Backend & Cloud:

DevOps, Hardware & Tools:

End-to-End Deployed ML Application

• Engineered advanced time-series features (rolling averages, lags) using Pandas and LightGBM to improve predictive accuracy by 66%.
• Integrated SHAP for model-agnostic explainability and containerized the application with Docker for serverless deployment on Google Cloud Run.

Offline Edge Inference & Knowledge Assistant

• Engineered a fully offline AI assistant using llama.cpp to run local LLMs (Qwen2.5-Coder) directly on hardware, eliminating cloud latency.
• Built automated scraping pipelines and integrated AnythingLLM for a local vector database, enabling offline Retrieval-Augmented Generation (RAG).

Distributed Multi-Agent Architecture

• Developed a distributed backend utilizing a master orchestrator to route complex KYC and credit modeling tasks to specialized worker agents using relational databases.

If you are impressed with my work, you can Buy me A Coffee: