Backend & Systems Engineer · Founding Engineer @ Nover
Distributed backends • Real-time streaming • GPU inference infra • DevOps & IaC
🔭 Founding Engineer & Tech Lead @ Nover — Owning end-to-end production infrastructure for an AI image generation platform: diffusion fine-tuning on H100s, ComfyUI inference serving on AWS, custom WebSocket APIs, and job orchestration. Took the product from initial development to public waitlist launch.
"I build the unglamorous half of a product."
Backend and distributed-systems engineer with founding-team experience. I own production infrastructure end to end — the APIs nobody sees, the queues that don't drop messages, the GPU pipelines that don't fall over at 3am, and the schemas that survive the next migration.
Backend is where I go deep. I'm comfortable full-stack enough to ship a working UI over my own APIs — I just optimize for what's behind them.
-
Today — Founding Engineer & Tech Lead at Nover. End-to-end engineering spanning the TypeScript/React application, real-time WebSocket services, backend APIs, job orchestration, and an autoscaled GPU inference platform on AWS. Fine-tuned diffusion models on H100s, implemented capacity planning and autoscaling, and shipped the public waitlist launch.
-
Recently — Backend & Data Automation at RankUp (EU, remote). Engineered a Topical Map Update System automating SEO keyword research and clustering with SBERT and DataForSEO, cutting manual strategy time by 40%. Built granular cost tracking (AWS S3) and structured logging (Rollbar) for real-time API spend visibility.
-
Before — Software Engineer at U LLC (Ohio, remote). Built a real-time EEG digital twin pipeline: streaming OpenBCI brainwave data, running Hilbert transforms on the wire, and feeding coupled oscillator models. Authored a Kuramoto-based EEG synchronization research paper; containerised backend services with Docker and enforced PyTest validation protocols.
-
On the side — Led Mind Cloud Robotics (70+ engineers) from a 9th-place ERC finish to 2nd and 3rd at UGVC, and migrated the entire stack to ROS 2 on Ubuntu 24.04 — a distributed systems problem dressed up as a robotics one.
What I ship: REST · gRPC · WebSocket systems · long-lived streaming connections · idempotent task APIs · microservices · event-driven backends · queue-based workloads · atomic task distribution · real-time stream processing · ETL pipelines · structured logging · container orchestration · IaC · CI/CD · GPU-backed service operation · live observability.
I lean on these when the system needs them — not the headline, but I've shipped real things with each.
AI Engineering & MLOps — diffusion models, GPU inference, LoRA fine-tuning, agentic systems
Frontend — enough to put a working UI over my own APIs
Robotics & Embedded Systems — SLAM, autonomous navigation, low-level I/O
Numerical & Analytics — when data needs shape before it ships
| Go Distributed Queue | Fault-tolerant task queue with atomic Redis ops, lease-based crash recovery, priority tiers, delayed scheduling, and dead-letter queues — validated with automated chaos testing. |
Go Redis Docker Prometheus |
|
| Agent Mesh | High-throughput orchestration engine sustaining 750+ RPS at sub-200ms latency via atomic Redis distribution, with a live React dashboard using temporal buffering for 60Hz rendering. |
Go React Redis WebSockets Docker |
|
| EnvGuardian | Encrypted .env Git workflow — age/SSH public keys, decrypt-compare write model, path-traversal hardened, fuzz-tested — backed by 8 ADRs and table-driven tests enforcing 80%+ coverage. |
Go age Cobra GoReleaser |
| Real-Time Arbitrage Engine | Four-service Docker Compose pipeline ingesting Coinbase/Binance WebSocket market data into Kafka, detecting BTC spreads with PyFlink tumbling windows, fanning events to Discord alerts and a Streamlit dashboard. |
Python Kafka Flink Docker |
|
| Agentic LLM Architectures | 12 agentic design patterns implemented as reusable LangGraph state machines — reflection, planning, tool-use, multi-agent orchestration — with Chroma retrieval, Tavily search, and LangSmith tracing. |
Python LangGraph LangChain Chroma |
|
| CerebralFlow | Neural dynamics framework — reconstruct functional networks from signals and run coupled-oscillator simulations with Hilbert transform analysis. |
Python SciPy NumPy |
|
| Physics-AI Bridge | Vectorized 2D Ising model (Metropolis-Hastings) with finite-size scaling, extended to Hopfield networks as learnable Ising systems for associative memory. |
Python NumPy |
| NetProbe | Cross-platform C++20 network traffic analyzer — live/offline capture, QUIC ClientHello SNI decryption, per-flow GeoIP/ASN, process attribution, and a Dear ImGui dashboard. |
C++20 Npcap Dear ImGui mbedTLS CMake |
|
| THAAD Simulation | 3D missile interceptor sim with Kalman-filtered radar tracking, augmented proportional navigation, realistic drag / G-limits, and interactive 3D replays with battle reports. |
Python NumPy Kalman Filter |
| Fourier Epicycle Machine | Draw any continuous path and watch a machine of rotating epicycles reconstruct it in real time — the DFT made tangible. Built with p5.js. |
JavaScript p5.js |
|
| dot-spend | Cross-platform, minimalist expense tracker in pure Python — ASCII analytics, CSV export, and status-bar integration for Linux ricing. |
Python CLI Linux |
| Nabda — Health Monitoring Platform | End-to-end cardiac-monitoring platform: a C++ wearable streams data into a Java / Spring Boot backend with PostgreSQL on RDS, Python services handle diagnostic reports and an AI chatbot, and Flutter + Next.js apps surface it to users. |
Spring Boot Java 21 C++ Python Flutter Next.js PostgreSQL Qwen3 |
yehiasaidgewily@gmail.com