Hamza Hasan Ellahie

• Built an automatic cv parsing pipeline using self-hosted oss models like qwen-7.5B, LLama3.1-8B, and Gemma3-12B
• Prepared hosting of models with auto scale to 0 to reduce cost while minimizing latency
• Deploy FastAPI backend for the parsing pipeline and provisioning of pods

• Developed a context-aware health platform using SwiftUI, HealthKit, CoreLocation and EventKit to correlate biometric heart data with user location and calendar events for health insights.
• Designed hybrid storage strategy for sqlite local caching and supabase for cloud syncs, ensuring 100% offline availability for health datasets.
• Implemented OAuth (Google, Apple), and leveraged swift concurrency for automated background data synchronization.

• Improved garbage collection in Earley parser, by applying graph theoretic and static-analysis techniques.
• Developed a custom test framework to benchmark multiple parsing and tree generation algorithms.
• Optimized memory usage in post-parsing tree generation from O(n) to O(1), resulting in up to 100x reduction on benchmarks

Implemented various Multi-Agent Pathfinding (MAPF) algorithms, scaling solutions from 4 to 50+ agents with around 40+% density Built a benchmarking suite to simulate real-world scenarios and measure algorithm performance and reliability. Profiled and implemented a parallelized solution, achieving up to a 65% faster execution on benchmarks Designed custom heuristics enabling around 25% of early pruning of suboptimal solutions in benchmarks.