Mohamed Deraz Nasr
About Me
GitHub Contributions
Loading contributions...
LeetCode Progress
Loading LeetCode heatmap...
Education & Experience
Education
BSc Software Engineering
University of Ottawa
GPA: 3.5
Experience
Machine Learning Researcher
Georgia Institute of Technology
- Researching neural decision trees for multimodal medical AI, producing concept-level reasoning on patient data (MIMIC-IV).
Machine Learning Engineer Intern
Shopify
- Prototyped PyTorch LLM to convert natural language to SQL, reducing querying time by 80% across 1M+ merchant data.
- Built ML pipelines (Python, SARIMA, Airflow, SQL) to forecast business metrics, improving accuracy by 7% across 5+ teams.
- Reduced model response time by 35% (Redis, Docker) to optimize real-time XGBoost predictions for merchant dashboards.
- Presented product to 20+ senior/staff engineers and team leads at cross-team review, earning highest performance grade.
Research Engineer Intern
University of Ottawa
- Researched distributed & reinforcement learning to accelerate solar energy forecasting and enable adaptive model updates.
- Engineered data pipelines (Python/Go) to process 200K+ solar readings daily with 80% accuracy via Redis caching.
- Deployed Next.js/GraphQL real-time dashboard with XGBoost forecasts to monitor model accuracy at <200 ms latency.
Full Stack Developer Intern
University of Ottawa
- Delivered data validation pipelines (Node.js/React Native) for 20K+ students, cutting form submission errors by 30%.
- Redesigned 50K+ record PostgreSQL database pipeline, reducing query latency by 40% under high concurrency.
Systems Engineer Intern
March Networks
- Built C/C++ debugging tool (Python/Docker) for RTSP/WebSocket on Linux, boosting model resilience by 25%.
- Developed Node.js/PostgreSQL dashboard to track packet loss and performance across 200+ deployed camera networks.
Extracurriculars
Embedded/Software Developer
UOBionics – Allonstride
- Engineered embedded C/C++ firmware (STM32/FreeRTOS) for exoskeleton motor & biosensor control (I2C/SPI/UART).
- Built (AWS/Python/WebSockets) pipelines with FastAPI/React dashboard for live anomaly detection on 100K+ signals.
My Projects
Affinity Map
Trained a few-shot protein embedding model with UMAP visualization to explore functional similarities across protein families.
RL Race Simulator
Designed multi-threaded race sim running 10K+ epochs, outperforming baseline strategies with 4x faster training throughput.
Protein Diffusion
Built diffusion-driven protein generator enabling fast 3D structure generation, conditioning, and biological plausibility scoring.
Financial Research Agent
Architected an agentic financial intelligence system using LangGraph and Gemini 2.0, achieving 100% data fidelity via SEC XBRL extraction and multi-tier cascading API architecture.
High-Performance Image Pipeline
Developed a multi-threaded C++17 image processing engine with O(N) separable filters and custom memory management for high-performance 4K video streams.
