Built web infrastructure for Abena AI, developed a cross-platform SDK ecosystem (React Native, Flutter, iOS, Android), and optimized model deployment pipeline with a 98% size reduction through advanced quantization.

Improved data-gathering systems and reporting tools, enhancing prediction accuracy by 80% using advanced statistical methods. Built data-driven web applications and consulted graduates on technical research papers.

Enhanced a biometric system using OCR and computer vision for multi-country compatibility. Improved facial recognition accuracy by 80% and engineered a scalable backend architecture for real-time analysis.

Co-founded a student/graduate job platform, leading full-stack development (Node.js, React, MongoDB) from MVP to production. Increased user retention by 40% through data-driven feature optimization.

A reinforcement learning agent trained using Group Relative Policy Optimization (GRPO) to efficiently navigate and solve complex mazes.

A C++-based noise cancellation engine that integrates classical Digital Signal Processing with a PyTorch-trained neural network to effectively suppress background noise.

A comprehensive gaze estimation system using morphological image processing for pupil detection and polynomial feature extraction with Ridge regression for calibrated screen coordinate mapping.

High-performance Breadth-First Search implementation leveraging Apple's Metal framework for massive GPU acceleration on large-scale graphs.

Built Nimble: a diffusion model that converted raster images to vectors with 85% shape retention. Trained on 10k+ samples in JAX.

A robust Rust library and command-line tool for parsing and evaluating mathematical expressions, supporting functions and scientific notation.

Client-side facial demographic analysis using WebAssembly and deep learning. Performs fast, private analysis directly in the browser with no data sent to servers.