Back to work
Case Study · Current Role
Baza Analytics
AI-powered analytics platform for non-technical teams. Multi-tenant workspaces, streaming LLM agents, human-in-the-loop data cleaning.
Role
Full-Stack Software Engineer
Timeframe
2025 - Present
Type
Product · AI Analytics
Overview
AI-powered analytics platform for non-technical teams. I work across the stack: shipped multi-tenant workspaces with RLS-backed data isolation, secure credential storage on GCP Secret Manager, five integration paths (Postgres, Gmail, Outlook, Google Sheets, Excel), a streaming Ask agent on LangGraph, and the Clean Agent - a human-in-the-loop data cleaning workflow with plan → execute → validate phases.
What I shipped
- Designed multi-tenant workspace model with Postgres RLS for data isolation across customer accounts.
- Built secure credential storage on GCP Secret Manager with workspace-scoped encryption envelopes.
- Shipped five integration paths (Postgres, Gmail, Outlook, Google Sheets, and Excel), each with its own auth, sync, and schema-inference path.
- Built the streaming Ask agent on LangGraph, including the retry/repair loop that catches malformed SQL before it reaches the user.
- Designed and shipped the Clean Agent, a plan, execute, validate workflow with human-in-the-loop approval gates.
- Debugged production issues across the stack: auth race conditions, RLS misses, LLM agent loops, streaming disconnects.
Stack
- Next.js 16
- React 19
- TypeScript
- FastAPI
- Python
- PostgreSQL
- LangGraph
- GCP Secret Manager