I’m Priyansh Goel. I build the retrieval infrastructure behind RAG and semantic search.

I build the retrieval infrastructure that decides whether a RAG or semantic search system actually works in production, and whether it makes money.

I’m a founding engineer at Sparkonomy, where I build production-scale multimodal semantic search across millions of minutes of video, image, and text. I own the retrieval pipeline end to end: ingestion, chunking, vector indexing, hybrid search, and re-ranking. I run large-scale vector retrieval in production and spend a lot of time on the recall-latency trade-off, the one that quietly decides your infra bill and whether the product feels fast enough to keep users.

Before Sparkonomy, I was Co-Founder and CEO of StepNex Technologies. I built scalable ERP systems for SMBs and ran the company end to end. That’s where I learned to read the business behind the build: unit economics, what customers actually pay for, and which technical bets compound instead of just looking good in a demo.

That’s the lens I bring to everything now. I think model selection is an infrastructure decision, not a vibe. I think your proprietary data layer is the only moat that compounds. And I think the boring parts of the stack, retrieval, evals, latency, are where the real leverage hides.

I work in Python, JavaScript, and Go, build on FastAPI, and live in vector databases, RAG, ANN search, and embeddings. 5x hackathon winner. B.Tech CSE from Manipal University Jaipur, class of 2026. Based in Gurugram.

I write about AI infrastructure, vertical AI, and model economics. If you’re building in that space, reach out.