Pensieve

Problem

Running multi-agent LLM workflows in production is hard to trust: outputs need human sign-off at the right moments, results must stream in real time, and LLM calls have to be routed and governed across providers for cost and reliability — without losing the audit trail.

Approach

I co-built and production-hardened Pensieve, an AI process-orchestration engine that runs multi-agent LLM workflows through human-in-the-loop approval gates, real-time streaming, and governed LLM routing across cloud providers. The architecture decouples orchestration from execution so agents, gates, and streams compose cleanly under concurrent load.

Impact

• Adopted by 2K+ users daily across domains.
• Human-in-the-loop approval gates make multi-agent output trustworthy in production.
• Governed, multi-cloud LLM routing for cost and reliability.

Contribution: Co-built and production-hardened — a collaborative effort I contributed to substantially.