Clinical AI Coding Service
Healthcare AI service suggesting ICD-10 and CPT codes from clinical notes via 3-layer hybrid retrieval — BioBERT semantic embeddings, BM25 lexical search, and historical co-occurrence — with clinical safety guardrails.
I design and develop high-performance web platforms, SaaS products, and AI-powered systems from concept to production.
I design and ship data-heavy systems — Django, FastAPI, ETL at scale, retrieval pipelines, and the production glue that keeps them up. Self-taught, came up through Lebanese agencies and freelance work, then went deep on backend and ML.
For the past year I've been the sole engineer at a healthcare technology company, where I've shipped two production platforms: a 714-model pharmaceutical reference data API and a clinical AI coding service combining BioBERT semantic embeddings, BM25, and historical retrieval.
Open to senior full-stack and AI engineering roles, and select freelance work on SaaS and AI-powered products.
Read more about meEnd-to-end product builds across backend, API, and frontend.
Production retrieval systems, LLM workflows, clinical safety guardrails.
Multi-tenant, multi-currency, multi-region — built to scale.
Healthcare AI service suggesting ICD-10 and CPT codes from clinical notes via 3-layer hybrid retrieval — BioBERT semantic embeddings, BM25 lexical search, and historical co-occurrence — with clinical safety guardrails.
714-model Django/DRF platform with a 309-CSV ETL pipeline, Redis cache warming, and 730+ versioned REST endpoints.
Custom React + Vite eCommerce platform powering multiple Lebanese retail brands — Garden State and LGT — from one codebase, with per-tenant config and feature modules.
Rebuilt a legacy desktop system on Electron + Angular + Node + Socket.io with offline-first cloud sync.
Multi-vendor, multi-currency eCommerce platforms for Kuwait — admin, vendor, integration, API, and customer surfaces.
Multilingual educational landing page with custom hand-coded animations. No template, no library — built from scratch as a craft exercise. Live at ata3allam.com.
End-to-end builds across Python, TypeScript, and the production layer between them.
Production retrieval systems, LLM workflows, multi-agent architectures, and the safety gates around them.
Multi-tenant, multi-currency, multi-region products — designed to scale before they need to.
REST and GraphQL APIs, versioning strategy, auth, rate limiting, structured logging.
Linux deployments, GitHub Actions, Docker, observability, accuracy-regression CI gates.
Translating ambiguous founder briefs into shipping product, with the engineering trade-offs surfaced early.
Click any row to expand. Most recent first.
Most AI portfolios stop at "I called the OpenAI API." Mine starts where that ends — with retrieval architecture, safety gates, evaluation pipelines, and the production glue that keeps an AI system honest in front of real users.
Map the real problem and constraints. Ask the questions the brief didn't answer.
Choose the boring, durable pieces. Sketch the data model, the API surface, the failure modes.
Ship vertical slices. Tests from day one. Structured logs from day one. CI from day one.
Tune the parts that actually need it. Caching, indexing, batching. Observability before optimization.
From SaaS platforms to AI-powered systems, I help turn ambitious ideas into production-ready software.