software engineer · backend & ai infrastructure
I'm Yehia Gewily, a backend and distributed-systems engineer in Alexandria. I work close to the path from user request to generated result: APIs, WebSockets, Redis job orchestration, GPU inference on AWS, storage, delivery, and release workflows. Around that work, I build queues, network tooling, simulations, and research systems in Go, C++, Python, and C.
brief
I'm Yehia, a backend and distributed-systems engineer and a founding engineer at Nover. I work across the path from an API request to a delivered AI result: real-time services, job orchestration, GPU inference, storage, deployment, and the failure handling between them.
Before Nover, I contributed to SEO data automation, EEG research software, and autonomous-rover systems. Together with my Electrical Engineering studies and the experience of helping lead a 70+ member robotics organization, that work taught me to value clear ownership, careful trade-offs, and software that stays understandable when something goes wrong.
previous work
Before Nover, I had the chance to contribute to data automation, computational-neuroscience research, and student robotics. Each role taught me something different about building carefully and working with people.
Backend & Data Automation Intern
I helped automate parts of an SEO research workflow, connecting FastAPI services with DataForSEO and language-model APIs. I also added cost tracking and structured logging so API usage and failures were easier to understand.
Software Engineer, R&D · Part-time
I worked on a real-time EEG pipeline for a non-invasive neuromodulation research effort. My contribution covered signal processing, a Kuramoto-based computational module, containerized backend services, and validation work for the research platform.
Volunteer · Chairman, Vice Chairman & Software Engineer
I joined as a rover software engineer and was later trusted to serve as Vice Chairman and Chairman of a 70+ member student engineering organization. In those roles, I helped align the software, electrical, and mechanical teams; supported the move to ROS 2; worked on university and industry partnerships; and coordinated funding, travel, and competition preparation. Together, the team earned second place at UGVC 2024 and third place at UGVC 2025. Earlier, I contributed to its ninth-place European Rover Challenge finish.
selected work
Commit your team's .env to git — encrypted — so cloning the repo is all it takes to have working config. age proves a file can be decrypted, never who sealed it, so every ciphertext also carries a detached signature that fails closed.
AI generation is distributed work: requests need state, queues, workers, progress, files, and failures users can understand.
L40S GPU inference backend Read case study →Workers crash, leases expire, and retries pile up. The queue makes ownership, recovery, and dead-letter behavior explicit.
35s max crash recovery Read case study →High event volume can make a live dashboard lie. This keeps task orchestration and browser rendering within honest limits.
750+ events/sec orchestration Read case study →Packet tools need capture, parsing, and rendering to stay decoupled while traffic keeps moving and protocols hide data.
60 tests · 3-platform CI Read case study →more projects
contact
Open to backend systems, AI infrastructure, and product engineering roles.