projects

Things I've built in labs, internships and my own startup: LLM pipelines, models of gene expression and immune binding, robots, and tools for designers.

1. Project PIE

   01
   

   A tool that reads new issues and pull requests in Keycloak and helps maintainers find the ones that need attention. Keycloak has more than 6,000 contributors and 2,500 open issues.

   GoGitHub APIPrompt EngineeringVector EmbeddingsRAGData Pipeline Architecture

2. Artery

   02
   

   A platform that helps designers at lifestyle hotels find vetted artists based on a project brief. I built the product, made the prototypes, and prepared the pitch. Microsoft for Startups backed the project.

   ReactNext.jsNode.jsTypeScriptPostgreSQLTailwindCSS

3. ORB

   03
   

   An embedded SLAM system that combines gyroscope and speed data to help autonomous space robots navigate, with a dashboard for monitoring operations.

   SLAMSensor FusionIMU IntegrationROS 2RTAB-MapPoint Cloud Processing

4. ProCapNet and BPNet

   04
   

   Bimodal deep learning models built on BPNet that predict transcription initiation from DNA sequence. The model identifies regulatory motifs and patterns and improved the lab's previous results by up to 15% in some cell types.

   Read the paper →

   PythonCNN DesignModel InterpretabilityDeepSHAPTF-MODISCOPRO-cap

5. pMHC-TCR Binding Model

   05
   

   Deep-learning methods that predict pMHC-TCR binding directly from molecular structure data, contributing to the lab's immune-response modeling research.

   PythonCNN DesignStatistical ModelingPCAModel InterpretabilityData Pipeline Architecture

6. Swatch

   06
   

   An automation platform for interior designers that pulls product data from vendor sites, organizes it into structured datasets, and generates client-ready slide decks in seconds, turning hours of data entry and formatting into a one-click workflow.

   Web ScrapingPlaywright AutomationPptxGenJSNode.jsTypeScriptData Pipeline Architecture