MR
SYS MR-2026 SENIOR GENAI & SYSTEMS ENGINEER
AVAILABLE

Projects

Things I've designed, built, and shipped.

All A2A / JSON-RPC AWS Anthropic Azure AD B2C Azure AI Search Azure OpenAI Claude Claude Code ClickHouse Cosmos DB Document Intelligence EKS FastAPI Gemini GitHub Actions Go Google ADK Jira LIT LangChain LangGraph Multi-tenant Next.js OpenAI PostgreSQL Python RAGAS Realtime voice WebSockets pgvector scikit-learn

SYS-01 · MULTI-AGENT REGULATORY COMPLIANCE AUTOMATION

RegLens

Multi-agent regulatory compliance automation. Feed it a regulatory PDF and your control matrix — a compliance research agent extracts every obligation, a gap analyzer checks each against your policies via RAG and scores the risk, and a report generator produces an audit report with a human-in-the-loop approval gate. Includes a drift-detection evaluation harness.

Orchestration
LangGraph supervisor + A2A/JSON-RPC agents
Models
Gemini (multimodal) + Claude (structured)
State
pgvector + Postgres-checkpointed, resumable
Deep dive → GitHub ↗ Live site ↗
PDF + control matrix a compliance run FastAPI REST · SSE progress stream LangGraph Supervisor ingest → retrieve policies → gap analysis → score risks → generate report → HITL · Postgres-checkpointed, resumable PIPELINE WORKERS Ingestion agent A2A · Gemini multimodal PDF → obligations pgvector RAG policy retrieval Claude gap analysis structured output Risk scorer A2A · Gemini + rubric gaps → risk levels HITL gate reviewer approves / edits / rejects A2A A2A PostgreSQL · LangGraph checkpoints · pgvector embeddings · run state · audit log CROSS-CUTTING A2A · JSON-RPC 2.0 evals · RAGAS · DeepEval · behavioral suite Presidio · PII guards OpenTelemetry · LangSmith · structlog Multi-agent · multi-model (Gemini + Claude) · human-in-the-loop · durable & resumable · Next.js UI · Docker / Terraform
Architecture — full drawing in the deep dive

  • SYS-02 · OPEN SOURCE · unscrew for spec ⟳

    Cloud Waste Hunter

    Cloud-agnostic resource monitor with ML-powered waste detection. Flags idle instances, unattached volumes, and stale snapshots across providers in a unified cost-optimization dashboard — then eliminates them safely with dry-run previews, human-in-the-loop approval, and 7-day rollback.

    STACK: Python · FastAPI · scikit-learn · Next.js
    Architecture
    Python · FastAPI · scikit-learn · Next.js
    DetectionIsolation Forest (idle EC2) + rules
    TargetsIdle EC2 · unattached EBS · old snapshots
    SafetyDry-run · approval · audit · 7-day rollback
    StackFastAPI · scikit-learn · boto3 · Next.js
    Deep dive → GitHub ↗