Unclaimed profileBuilt from public GitHub and Hugging Face activity. Michael has not joined Lounge yet.Claim it
Not on Lounge yet
Michael Legemah
developerSoftware & AI Engineer
Public projects
ml-fine-tuning-ai-engineering
—
HTML
langgraph-patterns
Three orchestration patterns as actually running code, not diagrams: supervisor/worker with real conditional routing, human-in-the-loop with genuine interrupt()/resume, and the same deterministic-first escalation algorithm as Sentinel. 19 tests, zero API key required.
Pythonagentagent-memoryagent-orchestration
rag-reference-pipeline
Reference RAG implementation tracing every choice to a numbered decision, not an unexamined default. Structure-aware chunking, provider-agnostic embeddings, hybrid dense+BM25 retrieval, negation-aware groundedness checking ported from Sentinel. 27 tests, zero API key required to run them.
Pythonbm25chunkingchunking-algorithm
quorel-free
Quorel is an MCP server that scores AI agent responses before they reach a user, a trust layer an agent calls mid-turn, not a dashboard someone checks after the fact.
TypeScriptai-observabilityevaluation-frameworkevaluation-metrics
liquid-capital
Multi-agent LangGraph treasury copilot. Four agents analyze liquidity and policy in parallel; a deterministic risk validator, not an LLM, decides if a transfer needs human sign-off before it can execute.
TypeScript
priauthra
Multi-agent LangGraph system for healthcare prior authorization. Checks eligibility, retrieves and cites real payer policy via RAG, drafts appeals on denial. Nothing reaches a payer without human sign-off.
Python
Likely stack
OpenAILangGraphMCP