AI Software Engineer specializing in agent systems, RAG, and production-grade product engineering.
I build AI systems that do real work: multi-agent pipelines, evaluation-aware retrieval apps, real-time collaborative tools, and full-stack products with solid backend, frontend, and deployment discipline. My work spans Python, TypeScript, Flutter, and core CS projects in compilers and systems programming.
- Procurement AI - 5-agent sourcing platform built with FastAPI, LangGraph, and cost-aware Claude routing for supplier discovery, verification, comparison, and outreach.
- Shipyard - coding-agent application with a persistent session model, explicit graph runtime, target-manager workflow, browser workbench, and long-run mission control.
- LegacyLens - production-focused RAG system for legacy Fortran and LAPACK code intelligence with evaluation and deployment scaffolding.
| Project | What I built | Why it matters |
|---|---|---|
| Procurement AI | AI-assisted sourcing platform with a 5-agent LangGraph pipeline, typed state, FastAPI backend, and Next.js frontend | Strongest signal for applied agent systems, model routing, and end-to-end AI product engineering |
| CollabBoard | Real-time collaborative whiteboard with AI board manipulation, Socket.IO sync, Firebase persistence, and LangSmith tracing | Demonstrates strong frontend systems work: 664 tests, sub-100ms sync, and 60 FPS canvas performance |
| LegacyLens | RAG workspace for scientific code intelligence with FastAPI, LangSmith middleware, eval scaffolding, and Railway/Vercel deployment | Strong proof of retrieval, evaluation discipline, and production-minded AI engineering |
| Shipyard | Coding-agent application with a persistent session model, typed tool layer, browser workbench, planner-backed execution flow, and long-run mission control | Relevant to AI SWE roles because it shows agent runtime design, tool orchestration, durable state, and operator-facing product thinking |
| Exquizite | Assessment platform that turns PDFs, Word docs, and spreadsheets into quizzes through async Bull and Redis pipelines | Strong applied-AI and backend signal: document processing, queue-based orchestration, real-time quiz sessions, and product delivery |
| MemoryManager | Custom C++ memory allocator with best-fit and worst-fit placement, hole coalescing, and sbrk()-backed allocation |
Cleaner systems signal than a coursework-style compiler because it shows low-level memory and OS-aware implementation work |
- End-to-end AI product engineering. I can design the model workflow and also build the APIs, UI, persistence, auth, and deployment around it.
- Full-stack system design. My work includes real-time collaboration, async processing, analytics, and production application flows.
- Strong fundamentals. Public work includes a custom memory allocator, async job systems, and event-driven architectures.
- Cross-platform execution. I have shipped across React, Next.js, Python backends, Node.js services, and Flutter mobile ecosystems.
AI: LangGraph, OpenAI, Anthropic Claude, RAG pipelines, evaluation workflows, multi-agent orchestration
Backend: Python, FastAPI, Node.js, Express, PostgreSQL, Supabase, Firebase, MongoDB
Frontend: React, Next.js, TypeScript, Flutter, Tailwind CSS, shadcn/ui
Data and infra: Redis, Bull queues, Socket.IO, WebSockets, Docker, LangSmith, Railway, Vercel
Foundations: compilers, memory management, Kafka, data structures, systems programming
B.S. Computer Science, University of Florida '23. Fluent in English and Arabic.
Other notable builds: Instad, PocketPay, kafka-tutorial, event-scribe-ai-assist, PLCProgrammingLanguage