A collection of research experiments, intelligent systems, computational work and open-source projects—exploring how ideas evolve from questions into working systems.
An active media-provenance system developed as part of OmniPulse, combining signed identifiers, signal-processing infrastructure and agentic orchestration to establish a verifiable write path for digital media.
An exploration of small language and scientific models through mathematical modeling, PDE-oriented computation and GPU-based experimentation, focusing on the relationship between model design and efficient computation.
Research exploring reinforcement-based optimization for latent reasoning, studying how representations and reasoning behaviour can be shaped through policy optimization rather than relying only on conventional token-level supervision.
A context-aware retrieval system designed to improve generation quality by combining semantic retrieval with metadata-aware document organization and contextual relevance scoring.
Find running shoes
An intelligent shopping agent built with LangGraph that turns natural-language recipes into personalized grocery lists, retrieves relevant ingredients and integrates with a Walmart-style shopping experience.
A Python package for visualizing neural networks as 2D diagrams and interactive 3D models.
A macOS MCP server that exposes Apple Reminders through a standardized interface, combining Python, FastMCP and AppleScript automation.
A personalized food recommendation system using KNN and TF-IDF based feature extraction, combined with external recipe and nutrition data.
The projects here are snapshots of a much larger process: learning, experimenting, publishing and building.