Moustafa Abdelhamid

A Next.js web app for a speech-to-text service. As the technical co-founder and full-stack developer, I built the entire application using a cutting-edge tech stack that included Next.js, TypeScript, Tailwind CSS, and Supabase. I helped the new startup run the entire stack with a GPU back end for transcription at a very low cost until we finished the MVP and got our first 10 happy customers.

An AI-powered meeting assistant that automatically joins video calls, records, and generates structured meeting notes. I built the full-stack application, including real-time bot orchestration, speech-to-text integration leveraging Transgate's transcription engine, and intelligent summarization using LLMs. The system handles meeting scheduling, automatic join, and post-meeting deliverables, including summaries, action items, and searchable transcripts.

I built and operated the AI platform behind E.ON's public-facing AI search product, serving millions of energy customers in Germany. I designed the RAG infrastructure: LangChain/LangGraph pipeline with Azure PostgreSQL vector search, full-text search, document ranking, deduplication, and query rewriting. I owned the full multi-tenant Kubernetes platform (dev/staging/prod on AKS) supporting multiple AI/ML teams, GPU-accelerated workloads, and production SLA. The platform hosted Ray Cluster, Prodigy, Langfuse, Argo Workflows, and more.

A fraud management system that detects suspicious transactions using machine learning and a rule-based engine. I worked in developing back-end services and MLOps that provide explained risk assessments of many types of transactions.

Noon is an eCommerce service based in Dubai and serves multiple Middle Eastern countries, and the main technologies I used included Scala, Play, Akka, Cassandra, CQRS, Protocol Buffers, and Mesos on AWS. I worked as a member of a big distributed team across three software companies to build software services for managing the company's warehouse.

A social network built in 2012 with many modern features where I designed most of the interactions to be real-time and worked on the main engine in both the back end and front end, contributing to developing the protocol and query language for real-time updates and the real-time engine in Scala that uses MongoDB as a datastore back end and WebSockets for sending real-time updates to the web and mobile clients in a way similar to Firebase.

A configurable virtual lab used for IT courses. After the trainees study the materials, they can start a lab with a specific scenario to apply what they learned. The lab consists of configurable virtual machines connected to a virtual network on AWS instances.

The students can interact with those machines through the browser that connects via VNC. We implemented support for Linux-based labs to start in a couple of seconds using Linux Containers (LXC) and Butter FS. The system consists of a cloud agent that manages running AWS instances and running containers and scaling up and down, the cloud agent that starts and stops containers on each VM, and guru-mate, a small program running on each container that checks the student's progress in the lab and gives them hints to proceed.

Built ETL pipelines and data infrastructure that supports analytics, search, machine learning, and other data science research and development. I have also developed an ML classifier using PyTorch to predict categories of educational videos.