Projects

Agents4Science — Multi-Agents for Scientific Discovery
🥇 1st place at the Cambridge–Infosys Hackathon 2026 · built on top of DenarioA literature-grounded peer reviewer for AI-generated science. Agents4Science scores a hypothesis on novelty, saturation, conflict, feasibility, impact, and evidence quality — then rewrites it into stronger variants and tells Denario which idea is actually worth writing a paper about.

Parallelization of HPC Algorithms
Centre for Scientific Computing, University of CambridgeA C++17 / MPI molecular-dynamics solver written for the Centre for Scientific Computing at the University of Cambridge. Velocity-Verlet integration for a 1D anharmonic oscillator and a 3D Lennard-Jones fluid, with a domain-decomposed parallel force computation.

DeepFlex AI Drug Design
Wells Wood Research GroupA temperature-aware deep learning model to predict residue-level protein flexibility (RMSF). Integrated ESM-C embeddings, 3D CNNs, and temperature-conditioned attention mechanism.

ClinicalAgents
Research Partner for CambridgeAn open-source multi-agent system designed to act as a research partner for clinical trial data. It mines academic papers and clinical trials, utilizing an adversarial 'Advocate vs Skeptic' debate loop to synthesize accurate, cited research briefs. Built in direct collaboration with biotech company researchers and academics.

Persona Ranker
Personal ProjectA high-performance lead ranking system that maps organizational hierarchies and scores buying intent using LLMs. Live Demo: persona-ranker-five.vercel.app · Documentation: persona-ranker-five.vercel.app/docs.

Digital Storage in DNA
Advanced Systems & Synthetic BiologyNovel methods for encoding vast amounts of digital data into synthetic DNA molecules. Implemented a 'Data Storage in DNA' project at UW.

New York Times Clone
Flutter Articles AppA Flutter project built using Clean Architecture to fetch New York Times articles, display article details, and provide more information via URLs. Inspired by Yamen Abdulrahman's work.

FactTrace Hackathon
🥉 3rd place at the Cambridge DIS Hackathon (Team Polaris)FactTrace: The Agentic Consensus Jury — determining whether an external claim is a faithful representation of an internal ground truth. Built for the Agentic Consensus Challenge, it detects distortions, exaggerations, and missing context, and explains its reasoning through a transparent, multi-agent process rather than a black-box True/False verdict. Developed by Felix Burton, Dhruv Gupta, and Uras Asil.


