I build governed agent systems — orchestration, gateway/HITL, enterprise RAG, loop engineering, eval contracts, and overnight autonomy — with 13 live demos you can inspect before we talk.
ai-architecture-portfolio — case studies, ADRs, trade-offs, and stack map (Nic Chin–style architecture hub).
→ Featured case studies & decisions · Live demos
vpeetla-ai-skills — org-wide agent skills for every repo: LangGraph, gateway/HITL, loop engineering, TDD, deploy.
git clone https://github.com/vpeetla-ai/vpeetla-ai-skills.git
./vpeetla-ai-skills/scripts/install.sh --cursor --codex --project .Inspired by mattpocock/skills + Karpathy agentic engineering + autoresearch.
Agent → Harness → Loops → Memory — LangGraph coding loop (Orchestrator · Review · Quality), ODAEU RAG tuning, real repo bug-fix → GitHub PR (never pushes to main).
▶ Live demo · API · LangGraph · MCP · Groq · GitHub PR workflow
Monitor → Govern → Remediate — AI Gateway for tool authorization: policy, HITL, signed audit, registry lifecycle, FinOps.
▶ Live control plane · Gateway SDK · OPA · FastAPI · Next.js
Principal-architect multi-agent OS — 3 LangGraph orchestrators, 7 RAG strategies (incl. Enterprise RAG adapter), loop patterns (ReAct · Reflection · Plan-Execute), gateway-wrapped notify.
Chief → parallel specialists → Critic → Slack / Telegram / WhatsApp · pairs with AegisAI
Multi-agent content pipeline — Research → 5 platform drafts → HITL → Publish.
▶ Live demo · LangGraph · Clerk · Render + Vercel
Governed knowledge layer — access-before-ranking, hybrid retrieval, AegisAI HITL bridge, Qdrant adapter, OTLP export.
Cross-repo regression contracts — versioned golden fixtures for RAG answers, LoopForge benchmarks, repo-fix, mission gates, and Content Factory HITL.
Validator · JSONL suites · CI · Evals-as-product proof
vLLM Architecture Lab · LIVE
PagedAttention · Continuous Batching · KV Cache — interactive 5-tab architecture explorer + Python engine simulator (scheduler, block allocator, memory budget API).
▶ Live demo · API · FastAPI · Educational simulator
Mission fleets + eval gates — FinOps estimates, Langfuse spans, VAP delegation, AegisAI human-gated ship, Netlify→FastAPI proxy.
Sentinel Brief · LIVE
Overnight AI radar — HN, arXiv, VentureBeat, MIT Tech Review, The Batch, Paper Digest, Towards Data Science, and more → snapshot diff → eval gate → governed email.
▶ Live demo · API · LangGraph · Resend · eval gate · cron
What should agents do? → Venkat AI Platform (orchestration)
What are agents allowed? → AegisAI (gateway, policy, HITL, audit)
What knowledge can they use? → Enterprise RAG (access-aware retrieval)
How do we operate them? → AegisLoop (missions, traces, eval gates)
What do they produce? → AI Content Factory (governed publish pipeline)
How do agents improve? → LoopForge (harness · LangGraph · repo fix → PR)
How do we serve LLMs? → vLLM Architecture Lab (PagedAttention · batching · KV budget)
How do we prove quality? → Golden Eval Registry (versioned eval contracts)
How do we scan overnight AI signal? → Sentinel Brief (daily governed intelligence brief)
| Layer | Repository | Live demo |
|---|---|---|
| Self-improvement | loop-engine-agent-platform | LoopForge UI · API |
| Agent governance | aegisai-enterprise-agent-platform | Control plane |
| Multi-agent OS | venkat-ai-platform | VAP console |
| Knowledge | enterprise_rag_platform | RAG lab |
| AgentOps | aegisloop-agentops-workbench | Mission console |
| Content automation | ai-content-factory | Content pipeline |
| LLM inference | vllm-architecture-lab | Architecture lab · API |
| Eval contracts | golden-eval-registry | Versioned golden eval fixtures |
| Overnight intelligence | sentinel-brief | Brief UI · API |
| # | Pattern | Repository | Live demo |
|---|---|---|---|
| 1 | ReAct | react-agent-pattern | Trace viewer |
| 2 | Reflection | reflection-agent-pattern | Trace viewer |
| 3 | Plan-Execute | plan-execute-agent-pattern | Trace viewer |
| 4 | Multi-Agent | multi-agent-system-pattern | Trace viewer |
| 5 | Swarm | swarm-agent-pattern | Trace viewer |
- LoopForge — LangGraph repo fix loop, PR workflow, live on Render + Vercel
- VAP ↔ AegisAI gateway (notify channels)
- VAP — 3 orchestrators + 7 RAG strategies + Enterprise RAG adapter
- Enterprise RAG ↔ AegisAI HITL bridge + Qdrant adapter
- AegisLoop — FinOps, Langfuse, VAP delegation, gateway ship path
- vLLM Architecture Lab — PagedAttention simulator live on Vercel + Render
- vpeetla-ai-skills — org-wide agent skills (Cursor + Codex)
- Golden Eval Registry — cross-repo regression contracts
- Sentinel Brief — governed overnight AI intelligence reporter live on Render + Vercel
- Live demos on Vercel for VAP, AegisLoop, Enterprise RAG, LoopForge, vLLM Lab, Sentinel Brief, and all 5 agent pattern repos
- Substack — venkatapeetla.substack.com — architecture decisions for production AI
- Medium — @vpeetla.ai — agent patterns & RAG governance
- Essay — From Multi-Agent OS to Agent Governance (VAP + AegisAI story)
- 🌐 venkat-ai.com
- 📰 Substack
If any repo helps you, a ⭐ helps other builders discover it.
