I build AI systems that ship to production. 12+ years of engineering experience spanning embedded systems, cloud infrastructure, and AI, currently part of the Data and AI team at Red Hat, focused on agentic AI, RAG pipelines, and LLM deployment.
- 💡 Patent Inventor: 10 invention disclosures and patents (1 issued + 9 filed) in generative AI, intelligent agents, and data security
- 🎤 Conference Speaker: Delivered talk at DevConf.IN 2026, Pune on production RAG systems; talk selected at DevConf.CZ 2026
- 🏆 Hackathon Finalist: Economic Times AI Hackathon 2026 (2nd Edition)
- 📝 Technical Writer: 30+ articles on Medium · 26K+ reads · Topics: RAG, agentic AI, LLM observability, transformers
| Project | Description | Link |
|---|---|---|
| Runkite | Open-source AI agent orchestration platform: self-hosted, framework-agnostic control plane for LangGraph, CrewAI, and custom agentic workflows | getrunkite.github.io/runkite |
| Chunking Strategy Library | Production-grade semantic text chunking with interactive visualization tool (thread-safe, streaming, adaptive retrieval feedback) | sharanharsoor.github.io/chunking |
| AntarDarshan | Production RAG system over 54 classical Indian philosophy texts using hybrid retrieval (BGE-M3 dense + sparse), cross-encoder reranker, Qdrant, FastAPI, Next.js (currently offline for maintenance) | GitHub |
Active contributor and maintainer of open-source AI/ML libraries. Created tools for agent orchestration, document processing, and LLM workflows adopted by engineering teams globally.
- chunking-strategy:
pip install chunking-strategy(Production-grade semantic text chunking) - Additional libraries: Agent orchestration, LLM memory systems, and more at github.com/sharanharsoor
Agentic AI & LLMs
ML / DL
Infrastructure
- Agent Skills: The Quiet Standard That's Changing How We Build AI Agents
- The Complete Guide to Embeddings and RAG: From Theory to Production (7.2K reads)
- Embeddings: A Deep Dive from Basics to Advanced Concepts (6K reads)
- The Rise of MCP: How a "USB-C for AI" Is Reshaping Intelligent Systems
- Understanding Mixture of Experts (MoE)
-
🎤 DevConf.IN 2026, Pune: "Why Your RAG System Hallucinates: Fixing the Content Segmentation Problem" · Watch on YouTube · Talk Details · Demonstrated 40-60% retrieval accuracy improvement live
-
🎤 DevConf.CZ 2026, Czech Republic: "Beyond Token Limits: Building Memory That Actually Works for LLM Agents" · Topic-aware LLM memory compression for agentic systems
Want to connect or book a mentoring session?
