I build production-grade GenAI systems — from vector-search infrastructure and multi-agent workflows to RAG pipelines and backend APIs. Most of my work is open-source and ships as real tools (PyPI packages, Docker-deployed services, CI/CD pipelines), not just notebooks.
I focus on making LLM applications fast (18.4ms p50 retrieval), reliable (0.92 MRR precision), and deployable (Docker + GitHub Actions + AWS).
| Metric | Result | Context |
|---|---|---|
| p50 Retrieval Latency | 18.4 ms | vectorDBpipe + ChromaDB |
| p99 Retrieval Latency | 115.8 ms | vectorDBpipe stress test |
| Ingestion Throughput | 42.5 chunks/s | vectorDBpipe modular pipeline |
| Retrieval Precision | 0.92 MRR | UPSC RAG engine |
| Ops Reduction | ~80% | Agentic workflow automation |
Open-source PyPI package that unifies FAISS, Pinecone, and ChromaDB into a single vector-search framework for LLM agents. Implements Omni-RAG, GraphRAG, and a Vectorless Engine. Ships as a real pip-installable package with a FastAPI dashboard.
Benchmarks: 18.4ms p50 latency · 0.92 MRR · 42.5 chunks/s ingestion
pip install vectordbpipe
Backend Repo · Frontend Repo · PyPI
Autonomous multi-agent system that reads emails, extracts tasks, generates smart replies, and maintains long-term memory via FAISS. Docker-deployed with Streamlit UI. Reduced manual extraction effort by ~80%.
Stack: CrewAI · LangGraph · FAISS · FastAPI · Docker · Streamlit · OpenRouter
RAG-powered chatbot for UPSC exam preparation using Gemini 2.0 Flash and Pinecone vector search. Implements conversational memory, dense retrieval, and context-aware response generation from uploaded documents.
Stack: Gemini API · LangChain · Pinecone · Flask · Python
Full-stack Java application that scans websites for WCAG accessibility compliance. Features automated rule checking, detailed violation reports, and remediation recommendations. Includes Jenkins CI pipeline and design documentation.
Stack: Java · Jenkins CI · Gradle · Backend + Frontend architecture
Collection of 5 production-oriented GenAI sub-projects: RAG chatbot with LangChain + ChromaDB, function-calling agent, LlamaIndex evaluation pipeline, multi-agent application, and transformer tokenization experiments.
Stack: LangChain · LlamaIndex · ChromaDB · Gemini API · CrewAI
Upload a document, get auto-generated MCQ quizzes with AI-graded answers. Uses Gemini Pro + LangChain for intelligent question generation with configurable difficulty and topic filtering. Deployed on GitHub Pages.
Stack: Gemini API · LangChain · Streamlit · Python
| Category | Repos | Highlights |
|---|---|---|
| AI / GenAI / LLM | 12 | vectorDBpipe, Agentic Workflow, RAG Chatbot, MCQ Generator, Blog Writer Agent |
| Testing & QA | 9 | Playwright automation, API frameworks, CI/CD pipelines, WCAG accessibility |
| Java & Backend | 5 | DrishyaScan, Ticket Booking System, Spring Boot, HTTP Server |
| ML / Deep Learning | 3 | COVID-19 CNN detection, Breast Cancer prediction, AI Core Foundation |
| DevOps / Infrastructure | 3 | Jenkins CI/CD, Terraform IaC, Git workflows |
| Web & Frontend | 6 | Portfolio site, HireSense-AI platform, production SaaS |
| Open Source | 1 | Hacktoberfest 2025 contributor |
- Published vectordbpipe on PyPI — installable via
pip install vectordbpipe - Contributed to BharatData — Indian public-data infrastructure (Python + TypeScript)
- Participated in Hacktoberfest 2025
- 38 public repositories across AI, backend, testing, and DevOps
- Active open-source contributor building tools, not just tutorials
