Specializing in LLM Orchestration, Multi-Agent Systems, and Cloud-Native AI Based in Hyderabad, India 🇮🇳

I build intelligent systems that don't just predict, but reason and execute. From local, air-gapped multi-agent frameworks for secure coding to multimodal RAG pipelines on GCP, I focus on turning experimental AI into production-grade infrastructure.

• tensorflow/tensorflow: Official Contributor to TensorFlow.
  • Engineered clearer diagnostic error messages for XLA compilation within tf.image.extract_patches.
  • Resolved cryptic TypeErrors by ensuring windowing parameters are validated as compile-time constants.
  • View my PR: tensorflow/tensorflow#108476
• GoogleCloudPlatform/python-docs-samples: Official Contributor to Google Cloud samples.
  • Engineered production-ready code for Vertex AI Agent Engine (Reasoning Engine) and Gemini AI integration.
  • Reliability Engineering: Resolved FileNotFoundError edge cases in image-generation samples by implementing automated directory creation and PEP 8-standardized path handling.
  • View my PR: GoogleCloudPlatform/python-docs-samples#13737

Production-ready, privacy-first RAG system for secure document intelligence.

• Impact: Built a high-performance local RAG pipeline that eliminates cloud dependencies, ensuring 100% data sovereignty for sensitive enterprise documents.
• Key Feature: Engineered a Semantic Chunking engine that splits text based on topic shifts rather than character counts, paired with LanceDB hybrid search for industry-leading retrieval accuracy.
• Tech: Next.js 15 (App Router), FastAPI, LanceDB (Vector Store), Ollama (Llama 3.2), Tailwind CSS.

Advanced Graph-RAG system for mapping complex global supply chain dependencies and ripple-effect analysis.

• Impact: Built a reasoning engine that traverses relationship nodes to identify hidden market risks, solving the "disconnected data" problem where standard RAG systems fail in multi-hop queries.
• Key Feature: Engineered a Hybrid Retrieval Pipeline that merges Neo4j graph traversals (Cypher) with ChromaDB semantic search, allowing the local LLM to reason across entities like suppliers, locations, and risk events.
• Tech: Python, Neo4j (Graph DB), Ollama (Llama 3), LangChain, ChromaDB, Streamlit.

Autonomous Multi-Agent System for Secure, Air-Gapped SDLC.

• Impact: Designed a local-first orchestration framework using LangGraph and Ollama to automate code generation, testing, and refactoring.
• Key Feature: Implemented a Self-Correction Loop where specialized agents (Architect, Coder, Reviewer) critique and fix code autonomously without cloud data leakage.
• Tech: Python, LangGraph, Ollama (Llama3/DeepSeek), Streamlit.

Agentic Drug Research Assistant powered by Gemini 2.0.

• Impact: Developed an intelligent agent that automates medical literature synthesis by combining Real-time Web Search with LLM reasoning.
• Key Feature: High-speed research synthesis using Gemini 2.0 Flash and DuckDuckGo integration, providing structured reports for healthcare professionals.
• Tech: Gemini 2.0 API, Python, Streamlit, DuckDuckGo Search API.

Real-time conversational AI with visual and auditory perception.

• Impact: Built a low-latency voice agent on Vertex AI capable of real-time screen processing and speech-to-speech interaction.
• Tech: Vertex AI (Gemini 1.5 Pro), WebRTC, Google Cloud Functions.

Time-series prediction for financial markets with 87% accuracy.

• Impact: Engineered ensemble ML models (LSTMs + XGBoost) for market trend prediction. Integrated data versioning and model monitoring.
• Tech: TensorFlow, Scikit-learn, Pandas, Time Series Analysis.

I'm looking for roles at the intersection of Generative AI, Finance, and Cloud Architecture. If you're building autonomous systems, let's collaborate.

💡 Learning Focus: Scalable Agentic Workflows and DeepSeek Model fine-tuning.