AI Engineer + Full-stack Engineer
Developer tools, agent workflows, APIs, and product interfaces—built end to end with TypeScript, Next.js, Python, FastAPI, and PostgreSQL.
01 / Selected full-stack AI work
AI agents, full-stack products, developer tooling, RAG, and production automation—engineered from interface to infrastructure.
Flagship project
Problem
Coding agents lose durable context, and opaque summaries make remembered claims difficult to trust.
Built and maintain a Git-native memory layer for coding agents. Evidence-backed claims, explicit provenance, append-only corrections, and repository validation make project knowledge durable, auditable, and safer to reuse across sessions.
Problem
Beauty brands need an elegant storefront that turns product discovery into effortless checkout.
Built and shipped a full-stack e-commerce experience for Irha Beauty with category browsing, product detail pages, cart, wishlist, and account flows, all in a refined nature-inspired design.
Problem
Service businesses lose ready-to-book leads when repetitive qualification waits on a human reply.
Built an applied AI workflow that responds to inbound duct-cleaning leads, gathers job details, checks fit and availability, and moves qualified conversations toward booking—with human takeover for exceptions.
Problem
Static learning content cannot answer follow-up questions without losing source context.
Built and deployed an interactive textbook with a RAG assistant that embeds and retrieves course content from Qdrant, returns source-grounded answers with lesson citations, and serves the experience through Docusaurus and FastAPI.
Problem
AI agent failures are difficult to reproduce, explain, and retest without replaying real-world side effects.
Built an agent observability and verification system that records execution, traces failures, and forks runs from the failing step with sandboxed tool calls to test a fix. It spans a Python SDK, FastAPI/PostgreSQL backend, and Next.js debugging dashboard.
More projects
Smaller tools and focused product experiments.
Built a full-stack AI task manager where authenticated users create, update, complete, and delete tasks through natural language. A Next.js client and FastAPI/PostgreSQL backend orchestrate an AI agent whose MCP tools execute scoped task actions.
An evidence-grounded Python CLI for company research, citation-validated drafting, and approval-gated outreach.
02 / Highlights
A compact view of the practice, projects, and systems I keep building in public.
10+contributions
4participations
3completed
BuildsProduction & open-source builds
AIAgents, LLM tooling & automation
StackNext.js, TypeScript, Python & APIs
OpenOpen-source engineering and knowledge sharing
NowGraphKeeper + Omniveer
03 / About
I work from system behavior and trust boundaries through APIs, data, interfaces, tests, and deployment.
I build AI systems and full-stack products where reliability, traceability, and human control matter.
My recent work includes grounded memory for coding agents, agent replay and evaluation, evidence-cited automation, and deployed web applications backed by APIs and PostgreSQL.
Founder of Omniveer.
04 / Core stack
A focused stack reflected in the projects above—not a catalogue of generic capabilities.
05 / Education