Jack Preuveneers

DPhil, Engineering Science

University of Oxford · Oxford, UK

• Computational fluid dynamics and numerical methods, sponsored by Dstl. Supervised by Luca Di Mare.
• Hudson Fellow: the sole Engineering Science DPhil admitted to St John's College in 2023, and the sole recipient of Hudson funding.

PGCert, Submarine Design and Acquisition

University College London · London, UK

• Distinction. Awarded best overall design and presentation by the Director of Submarine Acquisition.

MEng, Naval Architecture and Marine Engineering

University of Southampton · Southampton, UK

• Honours.
• Thesis: design of a ballast-free, zero-carbon LNG vessel.

AI Research Scientist (part-time)

Vela Partners · San Francisco, remote

• Build machine-learning systems and agentic research workflows for quantitative venture investing and startup evaluation.
• Published and patent-pending research on market quality, portfolio construction, and fund allocation.

Quantitative Analytics, VC portfolio construction

Oliver Wyman, via Vela Partners · Oxford, UK

• Developed statistical portfolio-construction models and presented quantitative strategy outputs to institutional limited partners.

Naval Architect

Ministry of Defence · Bristol, UK

• Led research into next-generation concept designs and prototyping tools for autonomous underwater vehicles and other platforms, improving design efficiency by more than 30%.
• Constructor Lieutenant Commander on the Trials Team, working with the Royal Corps of Naval Constructors on sea trials.
• Founded a national doctoral-training initiative, securing approximately £3.6m for a 30-student PhD cohort across UK universities.

Postgraduate Researcher

Shell · Southampton / London, UK

• Design and optimisation of V-shaped hydrogen carriers. The design was built as an £8,500 laboratory model and tested successfully.

Reasoning-Based AI for Startup Evaluation (R.A.I.S.E.)

Preprint · Vela Partners

• A memory-augmented, multi-step LLM decision framework combining reasoning-log distillation, structured rule extraction, ensemble candidate sampling, and persistent memory.
• Designed the quantitative evaluation for rare-event prediction; the combined pipeline improved precision by 54% and accuracy by 50% over the standalone-model baseline.

arXiv:2504.12090

Reasoned Rule Mining: precision-optimised weighted log-probability classification

IEEE CSCloud SecureFinAI, New York

• With Yigit Ihlamur. An interpretable LLM classification method using token log-probabilities, rule induction, logistic calibration, weighted log-odds fusion, and precision–recall-optimised decision thresholds. Patent pending.
• 24.5% precision against a 2% base rate, a 12.3x uplift, with a compute-aware cascade that removes 91.3% of the expensive LLM evaluations. Commendation Award, ICIM 2026, Oxford.

IEEE XploreWrite-up

High-efficiency swimming using hydrodynamic eigenmode decomposition

Intl. Conference on Computational Methods in Marine Engineering, Edinburgh

• With Amanda Smyth and Luca Di Mare. Solves generalised eigenvalue problems for unsteady potential flow, so that efficient kinematics can be identified directly rather than searched for.
• The full paper is in review at the Journal of Ocean Engineering and Marine Energy.

MARINE 2025

Quantitative Trading Simulation

Goldman Sachs × AmplifyME · London, UK

• Inaugural Oxford cohort. Ran market-neutral ETF arbitrage strategies on the MAG-7; placed 4th of 60 teams.

Algorithmic Trading Academy

Optiver × Oxford Alpha Fund · Oxford, UK

• Inaugural Oxford cohort. Built Black–Scholes market-making algorithms; reached the Varsity final, 6th of 80.

Trading Course and Challenge

Flow Traders · Amsterdam, NL