Computer Science engineer shipping production systems where drones, models, and microservices have to actually work — not just compile.
🇨🇾 Based in Nicosia, Cyprus · 🎓 University of Nicosia (Jun 2026) · ✉️ khadrahussein4@gmail.com · 💼 LinkedIn
name: Hussein Khadra
role: Software Engineer — AI/ML · Robotics · Backend
education: B.S. Computer Science & Engineering @ University of Nicosia (Jun 2026)
location: Nicosia, Cyprus
focus: Production-grade systems that bridge research and reality
stack: Linux + Docker + ROS 2 → TS/Node services → PyTorch models
currently: Hunting graduate / new-grad roles in applied ML & roboticsI work at the seam between applied ML and shipped infrastructure — the layer where a paper-grade idea has to survive containers, networks, noisy data, and a real user on the other end. Recent ground covered: a Cypriot-dialect ASR dataset built from scratch, an OCR preprocessing pipeline that handled documents engineers tried to feed it sideways, and a 5-UAV reinforcement-learned swarm I'm genuinely proud of.
Multi-UAV disaster-relief swarm. Five autonomous drones establish a self-healing WiFi mesh across collapsed infrastructure to restore connectivity for civilians and first responders.
What it does — Trained a PPO reinforcement-learning policy for swarm positioning against a log-distance path-loss network simulator, then dropped it into PX4 SITL + Gazebo Classic to see if the physics agreed with the math. They did.
Results
| Metric | Value |
|---|---|
| 🛰️ Mesh coverage on Baseline Controller | 96–99% |
| 📡 Aggregate throughput | 114.7 Mbps |
| ⚡ Latency | 6 ms |
| 🚁 Swarm size | 5 UAVs |
Stack — ROS 2 · PX4 SITL · Gazebo · PyTorch (PPO) · Docker · C++ · Python
End-to-end ETL producing a novel Cypriot-dialect speech corpus from broadcast TV, radio, and archival recordings — sourced, segmented, normalized, and formatted for ASR fine-tuning.
Designed reproducible audio preprocessing (format normalization, noise filtering, alignment) for dataset-scale model ingestion, and shipped the backend ingestion + storage for voiceofcyprus.org, the project's volunteer crowd-sourced annotation platform.
Stack — Python · Audio Preprocessing · ETL · ASR · Backend
🔗 (https://github.com/abasis-ltd)
Real-time CV pipeline: facial keypoint detection → 3D landmark projection → per-frame mask rendering. Optimized for low-latency video streams with angle and occlusion handling. The backend (ChickenVision Server 🐔☁️) handles the heavy lifting — frame processing, pose estimation, and the brain behind the cluck.
Stack — Python · OpenCV · Facial Keypoint Detection · pytorch-openpose
🔗 github.com/projectPavoculus/chickenvision-server · 🔗 pytorch-openpose
Production OCR preprocessing system using contour detection + Hough transforms + perspective warping. Engineered for the documents nobody wants — skewed, distorted, shadowed, photographed at 30° on a kitchen table.
Benchmarked third-party CV stacks, picked the winner, integrated into the production OCR path. Per-document processing overhead dropped, accuracy on the ugly inputs went up.
Stack — OpenCV · scikit-image · PyTesseract · Python
Production TypeScript/Node.js microservice for metadata extraction from documents and video, backed by Amazon S3 and consumed by the university student portal. Pluggable extractor architecture, Swagger UI'd, deployed at university scale.
Stack — TypeScript · Node.js · AWS S3 · Swagger · REST
Microservice worker for the Zelara project — plant species identification + care advice from image inputs. ML-backed classification stitched into a microservice architecture.
Stack — Python · ML Inference · Microservices
🔗 github.com/zelara-ai/zelara-plant
|
🧠 Also: PPO / Reinforcement Learning · ASR pipelines
🛸 Also: Gazebo Classic · PX4 SITL |
🌐 Also: REST APIs · Microservices
📊 Also: ETL pipelines · Audio + image preprocessing |
|
📍 MOST RECENT Abasis.AI · Nicosia, Cyprus
Built ETL pipelines producing a novel Cypriot-dialect speech corpus from broadcast TV, radio, and archival recordings. Designed reproducible audio preprocessing (format normalization, noise filtering, alignment) for dataset-scale ASR ingestion. Shipped the ingestion + storage backend for voiceofcyprus.org, the crowd-sourced annotation platform.
|
|
Goldman Solutions & Services · Nicosia, Cyprus
Built a document edge-detection pipeline using contour detection, Hough transform, and perspective warping — measurably lifting OCR accuracy on skewed and distorted scans. Benchmarked third-party CV libraries and integrated the chosen stack into the production OCR path, reducing per-document overhead.
|
|
University of Nicosia · Nicosia, Cyprus
Designed and deployed a TypeScript/Node.js microservice for metadata extraction from documents and video, backed by Amazon S3 and consumed by the university student portal. Delivered full-stack CRUD modules alongside senior engineers, integrated Swagger UI for live API documentation, and championed third-party libraries and AWS services adopted into production.
|
B.S. Computer Science & Engineering — University of Nicosia · Expected Jun 2026 · GPA 3.22 / 4.0 Data Structures · Database Management · Networks & Data Communication · Software Development · Discrete Mathematics · Linear Algebra
Certifications
- 🪪 IBM AI Applications with Python and Flask — Coursera (Nov 2024)
- 🪪 IBM Software Engineering Essentials — Coursera (Sep 2024)
Competitions & Hackathons
- 🏆 Bank of Cyprus Fintech Hackathon 6.0 (2025) & 5.0 (2024)
- 🏆 AI for the Common Good — British Council × Hack Cyprus (2023)
🛸 Iterating on multi-agent RL coordination beyond A.U.R.A
🎙️ Thinking about low-resource ASR for under-represented dialects
🧪 Building things that survive contact with production
☕ Open to graduate / new-grad roles in applied ML, robotics, or backend
If your stack involves drones, models, pipelines, or just well-engineered backends — I'd love to hear about it.
✉️ khadrahussein4@gmail.com · 💼 LinkedIn