CourseGen AI
AI-powered learning platform that curates and generates customized courses from open-source video content, intelligently ranked by quality using large language models.
I architect production-grade agentic platforms — multi-tenant, multi-region systems where agents plan, self-evaluate, and continuously improve. I run nda, a 10-agent development team that designs, builds, tests, and reviews code around the clock, and operate the full CredOS agentic pipeline in production.
The AI-Native Operating System for Intelligent Private Credit
Leading the architecture and delivery of CredOS — Penomo's agent-native operating system for infrastructure private credit. Architecting a self-owned six-layer backend: Hono + Vercel Fluid Compute API runtime, Temporal durable workflow engine, LangGraph + DSPy agentic core with GEPA self-improvement, LiteLLM multi-tenant gateway, and pgvector + Apache AGE — deployed across US and EU regions with full data residency. Owning the complete product lifecycle from customer discovery through technical architecture, eval harness design, and phased delivery.
Led development of Penomo's tokenized finance platform connecting global capital to renewable energy infrastructure. Architected smart contracts for real-world asset tokenization, built RESTful APIs for seamless blockchain integration, and engineered the complete frontend. Implemented KYC verification, real-time WebSocket communication, and cloud deployment on AWS.
Contributed to the development of Penomo's decentralized application, a platform enabling on-chain capital financing for real-world renewable energy infrastructure through tokenization. Built frontend components, integrated blockchain APIs, and worked with smart contracts in a production environment.
I'm an AI Solutions Engineer and Full Stack Developer who architects production-grade agentic platforms — multi-tenant, multi-region systems built to enterprise scale. I've designed and deployed 20+ AI agents, from durable workflow orchestrators and multi-tenant LLM gateways to citation-precise document intelligence, and run a 10-agent development team 24/7.
Currently leading platform architecture at Penomo, where I'm rebuilding CredOS — an agent-native operating system for infrastructure private credit — onto a self-owned six-layer backend. The stack spans Hono on Vercel Fluid Compute, Temporal durable workflow engine, LangGraph + DSPy agentic core, LiteLLM multi-tenant gateway, and pgvector + Apache AGE — deployed across US and EU regions with full data residency isolation.
I run nda — a coordinated 10-agent Claude Code development team with a Planner, Backend, Frontend, Tester, Security, Code-Reviewer, and more, each with persistent memory and cross-agent communication via a shared message board. Agent programs are built with LangGraph and DSPy, orchestrated by Temporal, with quality enforced by a Braintrust eval gate and full observability through Langfuse.
I run nda — a coordinated 10-agent Claude Code development team: a Planner that generates 3-plan proposals and opens PRs only after your approval, a Backend agent building Hono APIs and Temporal workflows, a Frontend agent writing React/TS, a Tester running Playwright Tier A + B suites, a Security agent auditing RLS and auth, and a Code-Reviewer keeping conventions clean. Each agent holds persistent memory, has a defined role, and cross-communicates through a shared message board and tiered memory store.
The CredOS pipeline never sleeps — Atlas extracts, Cipher classifies, Quill drafts IC memos with bbox-precise citations, all orchestrated by Temporal with zero silent failures. The GEPA optimiser self-improves DSPy programs weekly, with Braintrust eval gates and automatic rollback on regression.
Hono, React, Next.js, TypeScript. I build the APIs and interfaces that agents operate within — multi-tenant, multi-region, with WorkOS SSO, EU data residency, and tus multipart supporting 1 GB+ data rooms.
Not a single AI assistant — a coordinated engineering crew on Claude Code. The Planner never writes code: it orchestrates, generates 3-plan proposals, and opens PRs only after your approval. Every agent accumulates learned patterns across sessions.
Here are a few highlights. I've built 40+ projects across AI, automation, and full-stack development. Check out my GitHub for the full collection.
AI-powered learning platform that curates and generates customized courses from open-source video content, intelligently ranked by quality using large language models.
AI-powered SaaS platform that generates professional websites and intelligent voice agents for small businesses, complete with 24/7 AI receptionist, lead qualification, and appointment booking.
Chrome extension that analyzes job descriptions in real-time and generates tailored, professional cover letters using AI, directly from any job posting page.
Intelligent automation agent that securely stores user preferences and executes one-click purchases, streamlining the entire checkout experience.
Interested in AI-powered solutions, autonomous agents, or full stack development? I'd welcome the conversation.