Cool Projects
Spiking Neural Network Deployment Stack
Built a complete toolchain for deploying spiking neural networks to specialized hardware:
Core Runtime
Lightweight execution engine with optimized primitives for spike-based computation. Handles temporal dynamics and event-driven processing with minimal memory overhead.
Model Converter
Seamless integration with PyTorch and snnTorch for converting traditional deep networks to spiking architectures. Automated optimization passes for hardware constraints.
Visualization Tools
Interactive web-based tools for debugging and analyzing spiking network behavior. Real-time spike pattern visualization and performance profiling.
Contributed upstream improvements to open-source neuromorphic frameworks. Validated on event-based vision benchmarks achieving competitive accuracy with significant memory efficiency gains.
Comprehensive test suite with full coverage across all components.
Bizmila - AI Co-Founder Platform
Multi-agent LLM orchestration platform that transforms startup ideas into investor-ready deliverables.
Architecture: Led backend development using Node.js, TypeScript, FastAPI, and Docker.
AI Integration: Integrated OpenAI, Claude, and Cohere APIs with custom evaluation logic and intelligent fallback mechanisms across providers.
Workflows: Designed multi-agent workflows that handle ideation, market research, competitive analysis, and pitch deck generation.
Virtual Try-On Engine
Geometry-based try-on prototype with real-time computer vision and ML integration.
Backend Systems: Engineered the service orchestration layer and real-time data pipelines for processing video feeds and generating overlay predictions.