I build the tooling that lets AI agents ship real work — specified, isolated, reviewed, and merged.
And the agents themselves: autonomous digital employees that run without me.
| 🌍 100% Upstream Every merged PR landed in someone else's repository |
🏢 Trusted By GitHub · OpenAI · Pydantic CrewAI · Hugging Face |
🧩 One Ecosystem A full suite of composable spec-kit extensions |
Every badge above queries the GitHub API live — these numbers move without me touching this file.
 Pull Shark 
Merged pull requests at scale |
 Pair Extraordinaire 
Co-authored commits with others |
 Starstruck 
A repository the community starred |
 Quickdraw 
Closed an issue or PR within 5 minutes |
 YOLO 
Merged without review |
Awarded by GitHub for real contribution activity — not self-assigned.
I'm an Agentic AI Engineer working where large language models meet production software. My focus is the unglamorous half of agentic AI — the part that decides whether an autonomous system is trustworthy or merely impressive:
A specification the agent can be held to. Isolation so parallel agents don't collide. A checker that isn't the agent that wrote the code. And a stopping condition you can actually prove.
That belief turned into a body of work across four fronts:
| 🧩 Agent Tooling A suite of spec-kit extensions giving spec-driven development real teeth |
🏢 Digital FTEs Autonomous AI employees — scoped to a job, not a prompt |
👁️ Applied AI Computer vision, RAG pipelines, domain agents |
💻 Full-Stack Next.js products, mobile apps, hackathon builds |
role: Agentic AI Engineer · Full-Stack Developer
focus: Spec-Driven Development · Multi-Agent Systems · RAG · Computer Vision
building: the spec-kit extension ecosystem · autonomous digital FTEs
upstream: github/spec-kit · openai-agents-python · pydantic-ai · crewAI · smolagents
stack: Python · TypeScript · React / Next.js · FastAPI · Kotlin · OpenCV
open_to: open source collaboration · agentic AI roles · technical writingEvery merged pull request on this profile landed in a repository I do not own — maintained by GitHub, OpenAI, Pydantic, CrewAI and Hugging Face. Counts below are live.
| Project | Merged | Open | What I worked on |
|---|---|---|---|
| 🏆 github/spec-kit GitHub's spec-driven dev toolkit |
 |
 |
Dynamic expression engine for YAML specs, catalog/registry search, spec validation |
| 🤖 openai/openai-agents-python OpenAI Agents SDK |
 |
— | Agent tooling, runtime behaviour and developer-experience fixes |
| 🧬 pydantic/pydantic-ai Type-safe agent framework |
 |
— | Typed agent framework improvements |
| 🦾 openclaw/openclaw Autonomous agent runtime |
 |
— | Runtime behaviour for long-running autonomous sessions |
| 👥 crewAIInc/crewAI Multi-agent orchestration |
— |  |
Crew orchestration and agent-delegation improvements |
| 🤗 huggingface/smolagents Minimal agent primitives |
— |  |
Lightweight agent primitives |
| 🐍 openai/openai-python · panaversity | — |  |
SDK ergonomics · agent-factory business plugins |
A composable suite that takes spec-driven development from a document to an enforced workflow.
| 🔬 Analysis & Quality | |||
| spec-kit-brownfield |  |
 |
Retrofit spec-driven workflows onto an existing codebase — the hardest case, and the most useful |
| spec-kit-bugfix |  |
 |
Structured, reproducible bug-fix specifications |
| spec-kit-fix-findings |  |
 |
Automated analyze → fix → re-analyze loop — a closed loop with a real stopping condition |
| spec-kit-impact |  |
 |
Blast-radius analysis — see what breaks before you change a requirement |
🧪 Testing & Traceability | |||
| spec-kit-spectest |  |
 |
Auto-generate test scaffolds from acceptance criteria, map coverage, find untested requirements |
| spec-kit-trace |  |
 |
Requirement → test traceability matrix |
| spec-kit-ci-guard |  |
 |
Spec-compliance gates for CI/CD — verify specs exist, detect drift, block non-compliant merges |
⚙️ Orchestration & Workflow | |||
| spec-kit-worktree |  |
 |
Git-worktree isolation so parallel agents never overwrite each other |
| spec-kit-orchestrator |  |
 |
Cross-feature orchestration — track state, select tasks, detect conflicts |
| spec-kit-pr-bridge |  |
 |
Bridge specs to pull requests — the step that turns a plan into shipped work |
| spec-kit-branch-convention |  |
 |
Enforced branch naming across spec-driven work |
| spec-kit-api-evolve |  |
 |
Managed API contract evolution — breaking-change detection and semver guidance |
📐 Authoring, Scope & Cost | |||
| spec-kit-tinyspec |  |
 |
Minimal spec format for small, fast tasks — the on-ramp to the whole ecosystem |
| spec-kit-refine |  |
 |
Iterative refinement passes over a specification |
| spec-kit-scope |  |
 |
Effort estimation and scope tracking — estimate work, detect creep, budget time per phase |
| spec-kit-cost |  |
 |
Track real LLM dollar cost across SDD workflows — because token spend is the actual limit |
| spec-kit-diagram · changelog · toc-nav | + more | Diagram generation, changelog automation, navigation presets | |
⭐ Star and fork counters read GitHub directly — they climb on their own.
Beyond tooling: complete AI employees — agents scoped to a job description rather than a single prompt, with their own memory, tools and success criteria.
| 🎧 SaaS Customer Success FTE | An autonomous customer-success employee for a SaaS company — handles the recurring work a human CSM would own |
| 🎓 Course Companion FTE | A always-on educational companion that follows a learner through a full course |
| 🧑💼 Personal AI Employee | A general-purpose digital FTE — memory, tools and a standing job description |
| 🤖 Multi-Agent Todo Chatbot | End-to-end multi-agent architecture — conversational task management with agent handoffs |
⭐ open_ai_sdk |
Deep, hands-on exploration of the OpenAI Agents SDK — the groundwork behind my merged upstream PRs |
| Domain-specific autonomous agents — tool use, planning and multi-step reasoning across very different domains | |
| 📚 awesome-openclaw | Curated resource list for OpenClaw, the AI agent runtime I also contribute to upstream |
| ⏪ openclaw-rewind | Git-like time travel for AI conversations — rewind, branch and replay an agent session |
| 📖 Physical AI · RAG Humanoid Book · ai-native-book | Long-form technical writing on RAG, physical AI and AI-native development |
| 🍌 Nano Banana Prompt Tutorial · Context Engineering | Teaching material on prompt and context engineering |
✍️ Air-Writing System |
Write in mid-air using finger tracking and gesture recognition |
🎨 Gesture Drawing App |
Draw on screen with hand gestures — no mouse, no stylus |
| 🔊 Gesture Volume Control · 😀 Face Detection | Real-time hand-gesture system control and face detection pipelines |
| 🧮 Air Calculator | A calculator operated entirely through mid-air gestures |
📊 Crypto Portfolio Tracker |
Real-time portfolio tracking with live market data and analytics |
⚡ Gitup Profile Generator |
Generates rich GitHub profile READMEs from live account data — automation applied to developer identity |
| ✅ Full-Stack Todo App | Modern full-stack task management — TypeScript end to end |
| 🔐 Secure Data Vault · Password Generator | Encrypted storage and credential tooling |
| 📝 Blog Builder · 🧘 Zen Breathe | Content platform, and a minimalist mobile-first breathing app |
| 📱 FlightScry · 🏥 Medical Analysis · Health Care | Native Kotlin mobile app, plus health-data analysis tooling |
| 💼 Virtual Internships |
JPMorgan Chase & Co ·
TATA ·
BCGx ·
DATACOM
Industry simulation programmes — Java, Python and data analysis |
| 🥇 Hackathon Projects 13 repos · 60+ stars |
Marketplace 2025 ·
Furniture E-Commerce ·
Admin Panel ·
Documentation
Full e-commerce builds in Next.js + TypeScript, shipped under deadline |
| 🎨 Frontend & Portfolio | 3D Portfolio · Dynamic Resume · Coffee Site · Perfume Store · Parallax Scroll |
| 🎮 Games & Interactive | Adventure Game · Advanced Quiz · CLI Guessing Game · Cafe Game |
| 🏗️ Systems & Fundamentals |
Banking System ·
Library Management ·
Student Management ·
ATM Machine
OOP fundamentals applied to real domains |
📦 …and 100+ more repositories — Streamlit apps, Python utilities, CLI tools and coursework.
|
|
|
Also: advanced RAG pipelines · vector search (Qdrant / Pinecone) · context-retrieval design · custom evaluators & validators · LLM cost engineering
|
|
|
|
|
|
|
|
|
|
I'm open to open source collaboration, agentic AI engineering roles, and conversations about spec-driven development, multi-agent systems, and making autonomous work trustworthy.
💡 This profile maintains itself — every badge queries the GitHub API on view, and the activity feed refreshes on a schedule. Automation is the point.
