Projects
Published research is linked; the applied work is under wraps but described honestly. The tools run in your browser and are free to use.
Research
High-efficiency swimming from hydrodynamic eigenmodes
2025PaperEfficient swimming kinematics found by solving a generalised eigenvalue problem, rather than by searching the space of motions.
Unsteady potential flow around a deforming body, posed as a generalised eigenvalue problem whose eigenvectors are the modes of motion and whose eigenvalues carry the efficiency. Because the optimum falls out of the spectrum directly, the usual optimisation loop over kinematics disappears, and the eigenvectors turn out to obey a relation to their wave parameters that holds across planforms.
Fortran · Python · BLAS / LAPACK
R.A.I.S.E.: Reasoning-Based AI for Startup Evaluation
2025PaperA memory-augmented, multi-step LLM framework for predicting which early-stage companies become outliers, a rare-event problem with almost no labels.
Reasoning-log distillation, structured rule extraction, ensemble candidate sampling, and a persistent memory that carries what earlier decisions learned. I designed the quantitative evaluation; the combined pipeline improved precision by 54% and accuracy by 50% over the standalone-model baseline.
Python · LLMs · Rare-event evaluation
Reasoned Rule Mining
2025PaperA two-stage, precision-optimised LLM classifier for quant VC: 24.5% precision against a 2% base rate, a 12.3x uplift.
Interpretable classification from LLM token log-probabilities: induce rules, calibrate them, fuse the log-odds, and choose the threshold for precision rather than accuracy. A compute-aware cascade drops 91.3% of the expensive LLM evaluations before they run. With Yigit Ihlamur. Published at IEEE CSCloud SecureFinAI 2025, patent pending, and awarded a Commendation at ICIM 2026 in Oxford.
Python · LLMs · Calibration
Applied
Arbitrary-order DG solver for the 3D Euler equations
2024–PrivateA high-order discontinuous Galerkin solver for three-dimensional compressible flow on moving meshes, written in Fortran.
The numerical machinery underneath the doctorate: arbitrary polynomial order, moving-mesh handling for deforming geometries, and the performance discipline that comes with running it at scale.
Fortran · MPI · Numerical methods
VMQ: Venture Market Quality
2026PrivateA calibrated, explainable L1–L5 rating for early-stage venture markets. Describe a market in plain English; get a ranked rating you can argue with.
VMQ deliberately separates the model that predicts from the model that explains: an LLM analyst pipeline produces the rating, dimension scores, comparables and upgrade/downgrade conditions, while a quant model makes a real forward prediction scored against Crunchbase outcomes. Around that sits the investment workflow: population reassessment, a movers diff, value/momentum/quality lists, portfolio construction, and an expert review queue that doubles as an evaluation benchmark.
Flask · Next.js · BigQuery · Kronos
Vela Portfolio Optimiser
2025–2026PrivateMonte Carlo simulation and optimisation for venture portfolio construction: how many bets, how correlated, and what that does to the odds of holding zero outliers.
Correlated outcome sampling, Sharpe analysis, and constraint-driven optimisation over portfolio size and stage mix, ported from a desktop application to the web so the rest of the group can use it. Built for Oliver Wyman via Vela; the current work moves it off simulated distributions and onto real private-market data.
Next.js · Python · NumPy · Cloud Functions