Engineers, researchers, designers, and builders welcome.
First head-to-head empirical study of State-Space Models (Mamba) vs Transformers for autonomous music mixing. Learns tempo-coherent transition points from raw audio on MedleyDB a open research question.
Skills involved: Python, Machine Learning, LLMs, State-Space Models
Spaced repetition system that goes beyond SM-2 learning a personalised forgetting curve per card per user, predicting retention with a trainable model, and scheduling reviews to maximise long-term memory per unit time.
Skills involved: Python, Backend APIs, Machine Learning
Multi-perspective reasoning system fine-tuned on Reddit corpora that generates distinct supporting, factual, and opposing arguments probing whether LLMs can hold contradictory positions simultaneously.
Skills involved: Python, Backend APIs, LLMs
Self-supervised audio feature extractor and cross-platform recommendation engine. Learns latent music representations via contrastive learning, then retrieves similar tracks using approximate nearest-neighbour search.
Skills involved: Python, Machine Learning
LLM-native document assistant that streams context-aware autocomplete, runs ideation and critique loops inline, and auto-generates structured charts turning Docs into an active thinking partner rather than a passive editor.
Skills involved: Python
Tool-augmented LLM planner that decomposes natural language travel goals into multi-step itineraries, grounds each recommendation with live API data, and resolves constraint conflicts across days and budgets.
Skills involved: Python, LLMs
Voice-first personal knowledge system: Whisper transcribes speech in real time, an LLM pipeline restructures and categorises the output, then stores semantically indexed notes making spoken thought instantly searchable.
Skills involved: Python, Backend APIs, Docker, Databases
ML-driven prediction market trading system that fuses ensemble models, Kelly-optimal sizing, and walk-forward validation to exploit mispricings in real-money crypto markets 127% return, 3.15 Sharpe ratio.
Skills involved: Python, Docker, Machine Learning, Data Science, Quantitative Finance
Multi-agent LLM system where persistent AI characters with individual memory, goals, and social relationships generate emergent school narratives a testbed for long-horizon coherent agent behaviour.
Skills involved: Python, Backend APIs, Game Dev
Production-grade card game engine with a full PokerHandEvaluator, 6 complete games, and deterministic RNG designed as a clean substrate for training GTO poker agents and game-theoretic AI research.
Skills involved: Python, Game Dev
Texas Hold'em agent targeting GTO equilibrium play via Deep Counterfactual Regret Minimisation the same algorithmic family behind Libratus and Pluribus, the first systems to beat professional players.
Skills involved: Python, Game Dev
WWII grand strategy simulation where 95 nations are live LLM agents with historically-grounded personas — they negotiate alliances, declare wars, and respond to diplomatic overtures in real time via SSE streaming.
Skills involved: Python, TypeScript/React, Backend APIs, Game Dev
Agentic MUN preparation system: RAG-based researcher builds position papers from live sources, an LLM judge scores speeches against delegate criteria, and a debate simulator runs adversarial argument rounds.
Skills involved: Python
Bidirectional BFS traversal of Wikipedia's 60M-edge hyperlink graph to find minimum semantic separation between concepts a concrete probe into how human knowledge is topologically organised.
Skills involved: Python
Predictive FPL engine that reinvents xG for fantasy scoring, models fixture difficulty with Bayesian updating, and frames optimal weekly transfers as a constrained optimisation problem beating naive heuristic strategies.
Skills involved: Python, Data Science
Temporal social graph analyser for group chats tracks evolving relationship dynamics, detects influence patterns, and applies LLM summarisation to surface latent themes across thousands of messages.
Skills involved: Python, LLMs
Cross-modal popularity signal aggregator that fuses LLM sentiment, engagement velocity, and media frequency into a unified real-time metric modelling the information cascade that drives public attention.
Skills involved: Python, LLMs
Multimodal personality profiler that fuses webcam micro-expression analysis, prosodic voice features, and situational response embeddings moving beyond self-report bias toward behavioural ground truth.
Skills involved: Python
Ground-up social platform with structured debate threads architected to study how argument topology and community norms shape opinion dynamics. A research substrate for computational social science.
Skills involved: Python
A structured ethical reasoning framework for dual-use technology decisions scoring projects across capability overhang, misuse asymmetry, and counterfactual impact. Relevant as AI capabilities accelerate.
Multi-tracker video labelling suite with CSRT/KCF/TLD support built as a platform for generating high-quality training data for foundational vision models, where annotation throughput directly bottlenecks model quality.
Skills involved: Python, Computer Vision
Don’t see the right project?
Just reach out.
If nothing above is quite right but you’re excited about the kind of work I do AI, sports, trading, research just say hi. The best collaborations often don’t start with a perfect brief.
Tell me what you’re good at, what you’re curious about, and what you’d love to build. We’ll figure it out from there.
Start a conversation