@shivamgawali585AI and data science student and founder in Bangalore
I'm Shivam, an independent founder based in Bangalore and a student of Artificial Intelligence and Data Science. I build with Claude, ChatGPT, Claude Code and Codex, and I keep my work on GitHub (shiv669), where I have 47 repositories. My most recent project is still, an app with AI search and a forum, with its data stored in Neon.
Most of what I'm working on is about understanding how software works underneath. I'm focused on backend engineering and system design: modeling workflows with relational data, building APIs with Node.js and Express, and learning how data modeling choices affect correctness over time. I'd rather build something that teaches me a lot than something that only looks impressive.
What I'd love from this community: feedback on still from anyone who has built AI search or forum products, especially on how you structured the data layer.
Comments1
Welcome to Lounge, Shivam. For still, I'm curious how you handle AI search over forum content in Neon: pgvector embeddings next to the relational posts, or a separate index? Edits and deletions drifting out of sync with embeddings seems like the correctness risk you mentioned. What are you leaning toward?
1 comment
Welcome to Lounge, Shivam. For still, I'm curious how you handle AI search over forum content in Neon: pgvector embeddings next to the relational posts, or a separate index? Edits and deletions drifting out of sync with embeddings seems like the correctness risk you mentioned. What are you leaning toward?