Skip to content

I build AI
AI systems that scale.

Software engineer specializing in AI-powered solutions, RAG systems, and automation. Building scalable products at TJM Labs across healthcare and customer service.

View ProjectsGet in TouchView Resume

About

I'm a software engineer based in Cairo, Egypt. I specialize in AI-powered solutions, RAG systems, and robotic process automation for healthcare and customer service.

I joined TJM Labs as the second engineer when the company was just three people. Since then, we've scaled into a large team, raised $100M, grown to 450+ live pharmacies, and made 2 acquisitions, with 500K+ tasks now automated daily. I lead delivery of 20+ automation projects and manage the RPA development team.

Outside of work, I've competed in international challenges like the ALLaM Challenge (17 countries, 177 teams) and NASA Space Apps. I ship side projects too: an open-source MCP server that cuts an agent's always-on tool cost by more than half without losing a single capability, and a live Egyptian market dashboard that now has an Android app on Google Play.

30+

Projects managed

Nov 2024

Engineering since

Skills

Technologies I work with

AI & Machine Learning
  • Python / PyTorch
  • LangChain / RAG Systems
  • MCP (Model Context Protocol) / Agent Tooling
  • Computer Vision (YOLOv8, GANs, Stable Diffusion)
  • IBM Watsonx / LoRA Fine-tuning
  • Elasticsearch / Vector Search
Backend & Automation
  • Django / FastAPI / Fastify
  • Go / Node.js / Express
  • PostgreSQL (Drizzle) / SQLite / MongoDB
  • Docker / Docker Compose
  • RPA (BotCity / Playwright) / Retell AI
Frontend & Mobile
  • Next.js / React / TypeScript
  • Flutter / Riverpod
  • Tailwind CSS
  • Arabic RTL / i18n (next-intl)
  • PWA / Trusted Web Activity
DevOps & Tools
  • CI/CD / GitHub Actions
  • Hetzner / Vercel / Cloudflare
  • Linux / Nginx
  • AWS Lambda
  • Git / GitHub

Experience

Where I've worked

TJM Labs

Nov 2024 - Present

Joined as the 2nd engineer. Company scaled into a large team, raised $100M, grew to 450+ live pharmacies, made 2 acquisitions, and now automates 500K+ tasks daily.

Software Engineering Lead

Sep 2025 - Present
  • Leading delivery of 20+ automation and AI projects across healthcare clients including Gifthealth, where TJM Labs automation drives $25M+ in annual savings and 25x ROI across 25,000+ prescriptions per day
  • Leading engineering across 3 product divisions (Retail, Voice AI, and Long-Term Care) with a team of 8 engineers in Egypt, coordinating with division managers on resource allocation, code reviews, and hiring, and contributing to the company's internal technical direction
  • Owning infrastructure and internal tooling: provisioned and operate the Hetzner VPS fleet, deployed and maintain TJM Connect (self-hosted RustDesk for remote support)Read case study
PythonDjangoTeam LeadershipAIRPAHealthcare

Software Engineer

Nov 2024 - Sep 2025
  • Contributed to ColiVara, TJM Labs' open-source visual document retrieval platform: built the ColiVara-Eval benchmarking suite, Python and TypeScript SDKs via OpenAPI Generator (saving $30K+/year on SDK maintenance), and a public HuggingFace demo space
  • Built an end-to-end voice customer service system (Retell AI + Django) handling 50% of inbound client phone calls in production
  • Built and deployed production RPA automation bots (BotCity + Playwright) for healthcare clients, automating high-volume pharmacy workflows including prescription intake, fax processing, and patient demographic updates
PythonDjangoLangChainRAGRPABotCityRetell AI

Data Science Supervisor

IEEE Mansoura CSC

Dec 2023 - Oct 2024Volunteer
  • Developed advanced machine learning curriculum covering supervised/unsupervised learning, deep learning, and NLP
  • Mentored and trained the next generation of data science committee members
  • Led technical workshops and hands-on coding sessions
PythonMachine LearningData ScienceMentoring

Education

Academic background

B.Sc. Computer and Control Systems Engineering

Mansoura University

2020 - 2025
  • Cumulative academic total: 91.8%
  • Languages: Arabic (native), English (professional)

Trained ML models to predict civil engineering structural behavior, achieving R-squared of 96.5% - surpassing the original finite element analysis benchmark of 91%.

Training and Specialized Programs

WorldQuant University, DeepLearning.AI, ITI

2023 - 2024
  • WorldQuant University - Applied AI Lab: Deep Learning for Computer Vision (6 PyTorch capstone projects covering YOLOv8 object detection, MTCNN face recognition, GAN medical imaging, and Stable Diffusion deployment)
  • WorldQuant University - Applied Data Science Lab (8 applied data science projects)
  • DeepLearning.AI - Deep Learning Specialization (5-course Coursera specialization, ACE-endorsed: CNNs, RNNs, LSTMs, Transformers, Attention, TensorFlow)
  • ITI (Information Technology Institute, Egypt) - Computer Vision intensive (60 training hours, government-funded ICT ministry program)

Projects

Things I've built

Featured

Full Stack Android App

Alban Osama

Production Flutter + Fastify app I designed, built, and shipped solo for the dairy section of my father's supermarket in Egypt. Arabic RTL, Google Sign-In with admin approval, multi product bookings, live chat, counter offers, PDF delivery cycle reports. Live on Google Play.

FlutterRiverpodGoRouterTypeScriptFastifyPostgreSQLDrizzle ORMFirebase FCMWebSocketsPDFKitDockerHetznerGitHub Actions
Play StoreView PrivateDetails
Featured

Open Source - MCP Server

WhatsApp MCP Server

Self-hosted, authenticated MCP server over my own WhatsApp history. Serving the model 29 of its 96 tools instead of all 96 made it pick the right one more often, not less, at roughly 56% less always-on context. There is a labeled eval in the repo that measures it. Go bridge + Python FastMCP, MIT.

GoPythonMCPFastMCPwhatsmeowSQLite (FTS5)DockernginxPrometheusHetzner
GitHub Details

RAG Platform - Contributor

ColiVara

Contributor to TJM Labs' visual document retrieval platform (1.4k+ GitHub stars). Built the ColiVara-Eval benchmark harness, Python + TypeScript SDKs via OpenAPI Generator, and a public HuggingFace demo space.

PythonTypeScriptRAGVision EmbeddingsOpenAPI GeneratorBenchmarking (NDCG@5)LangChainDocker
View Details

AI Platform

AQSA

Arabic Quality and Skills Accelerator, built with a team for the ALLaM Challenge 2024 (17 countries, 177 teams). Six AI features on IBM Watsonx + ALLaM. I led the Quran analysis, Quran search engine, and Mo3gam dictionary features.

IBM WatsonxALLaM LLMLangChainPythonFastAPIElasticsearchFlutterTypeScript
PrivateDetails
Featured

Full-Stack App

Finance Dashboard

Egyptian market dashboard: gold, silver, USD parallel and official, forex, crypto, inflation. Arabic first and RTL by default. Rebuilt from a vanilla JS SPA into a Next.js 16 + Postgres monorepo with append only price history, and now ships to Google Play as an Android app in closed testing.

Next.js 16TypeScriptPostgreSQL 16Drizzle ORMTailwind v4next-intl (RTL)TradingView ChartsPlaywrightVitestsystemdCloudflareHetznerTWA (Google Play)
View Details

Deep Learning

Brain Tumor Classifier

Custom PyTorch CNN classifying brain MRI into four tumor categories (no tumor, pituitary, glioma, meningioma) at 99.3% test accuracy. Shipped as a live Streamlit demo so anyone can upload an MRI and get a prediction in the browser.

PyTorchCNNComputer VisionBatch NormalizationStreamlitJupyter
View GitHub Details

Neural Network from Scratch

AI Cars (Vanilla JS)

Self-driving cars trained with a genetic algorithm. Hand-rolled neural network (5 sensor rays -> 6 hidden -> 4 outputs), elitism, adaptive mutation when training plateaus, live fitness chart, network visualization, mobile touch D-pad. Pure vanilla JS, zero dependencies, no build step.

Vanilla JSNeural NetworkGenetic AlgorithmHTML5 CanvasZero Dependencies
View GitHub Details

Browser Games

Games Hub

Gallery of browser games, each on its own subdomain. Codenames (EN + AR, 500+ words per language, full rules enforcement) and a classic Snake with a live leaderboard API. Zero frameworks, vanilla JS, HTML validated in CI.

Node.jsExpressWebSocketsHTML5 CanvasVanilla JShtml-validate
View Details

Contact

Let's connect

I'm always open to discussing new projects, interesting ideas, or opportunities to collaborate. Whether it's AI, automation, or building something from scratch - drop me a message.

View ResumeDownload PDF