Research

At the Frontier of
Finance and AI

Active in ML, RL,
and LLM research
for trading and finance.

Medes Capital is an active research organisation operating at the intersection of machine learning, reinforcement learning, and quantitative finance. Our work spans the full spectrum — from foundational model research to applied strategy development.

We study financial markets as complex, adaptive systems. Our research applies the most current advances in artificial intelligence to problems of prediction, execution, risk, and signal discovery — building models that are not just theoretically sound, but empirically validated.

01

Machine Learning
for Trading

We develop and evaluate supervised and unsupervised machine learning models for market prediction, factor modelling, and cross-asset signal generation. Our focus is on methods that are robust to regime changes and non-stationarity — the defining challenges of financial data.

Deep Learning Gradient Boosting Regime Detection

02

Reinforcement Learning
for Finance

We train adaptive agents that learn optimal execution and allocation policies directly from market interaction. Our RL research addresses the unique challenges of financial environments: sparse rewards, transaction costs, and adversarial dynamics.

Policy Optimisation Execution Portfolio RL

03

Large Language Models
in Finance

We actively research LLMs for financial applications: parsing earnings calls, extracting alpha signals from news and regulatory filings, and building domain-adapted language models for quantitative tasks. We study both proprietary model fine-tuning and zero-shot financial reasoning.

Sentiment Analysis Information Extraction Fine-tuning

We aim to
publish our work
to help others.

We believe that advancing the state of the art benefits everyone — researchers, practitioners, and the broader financial ecosystem. Where our work is not commercially sensitive, we intend to share findings through academic publications, technical papers, and open contributions to the research community.

Our researchers are encouraged to engage with the wider academic world, participate in conferences, and contribute to the growing body of knowledge at the intersection of AI and quantitative finance.

Publishing is not just an obligation — it is a commitment to the discipline we have built our work around.

Proprietary Research

Most of our research is conducted in-corporation.

The majority of Medes Capital's research is proprietary and remains confidential within the organisation. Strategy development, live signal research, model architectures, and trading system designs are protected intellectual property and are not disclosed publicly. We share what benefits the community — and protect what drives our edge.