I build AI systems that work in production — RAG pipelines, LLM agents, and voice interfaces that handle real load and deliver measurable results.
Most of my work is hands-on: real retrieval architectures, agentic workflows, and backend systems built to scale. I care about shipping things that actually run — not just notebooks that demo well.
I contribute to open source, intern at AI startups, and spend a lot of time thinking about how to make LLMs faster, cheaper, and more accurate in the real world.
- RAG & GraphRAG — hybrid vector-graph retrieval using FAISS, Neo4j, and LangChain that meaningfully improves retrieval speed and answer accuracy
- Agentic AI — multi-agent orchestration with LangGraph, tool-calling pipelines, and structured prompt optimization
- Voice AI — end-to-end voice pipelines with Whisper + Ollama achieving sub-2s response latency
- Backend systems — FastAPI microservices, Dockerized deployments, and REST APIs built for real concurrency
- Automation — scripting workflows, CI/CD pipelines, and reducing manual overhead wherever possible
- AI Developer at Ekthaa — deployed a production RAG chatbot serving 500+ businesses, improving retrieval speed by 70% using FAISS, Neo4j, and FastAPI
- Former AI Engineering Intern at Regality AI — improved multi-hop QA accuracy by 35% and cut LLM deployment time by 60% using GraphRAG and Dockerized microservices
- Open source contributor — contributed agent collaboration modules to Hive, a swarm intelligence framework for distributed multi-agent AI systems
- B.Tech in CS (AI & ML) at Sreyas Institute of Engineering and Technology, graduating 2026
Building in AI? I'm always up for a conversation — reach out.
| Domain | Technologies |
|---|---|
| AI / ML | LLMs, RAG, GraphRAG, LangChain, LangGraph, Ollama, FAISS, PyTorch, NLP, Prompt Engineering |
| Backend | Python, FastAPI, REST APIs, JWT Authentication |
| Databases | Neo4j, PostgreSQL, MongoDB, MySQL |
| Cloud / DevOps | Docker, AWS, Git, Linux, CI/CD, Microservices |
- Campus Eats — Full-stack ordering platform supporting 200 concurrent users with real-time order tracking via FastAPI, PostgreSQL, and Redis
- AI Resume Screener — NLP pipeline using TF-IDF and cosine similarity that cuts manual screening time by 60%
- Voice-Activated AI Assistant — Modular Whisper + Ollama pipeline with sub-3s end-to-end response time and real-time web retrieval
- Email: shiva24.santosh@gmail.com
- LinkedIn: Shiva Santosh Reddy Aenugu
- Portfolio: shiva9198.github.io