AI4Finance Lab Contact
A virtual research lab for AI in finance

What does AI really add to finance when it is tested honestly?

AI4Finance Lab studies how AI systems use financial evidence, how they should be designed, and how their results should be evaluated so that they hold up outside the lab.

Explore our research Publications

Three kinds of evidence. One coordinated view. An honest test.

Our mission in motion
Market regime · shared context Fundamental Technical Text and sentiment Coordinator agent weighs, arbitrates, may abstain known at t One forecast, with its uncertainty Tested walk-forward, net of costs train → predict the unseen future → roll forward
  1. 01
    Specialists read the evidence
    Fundamental, technical and text signals, each analysed on its own terms.
  2. 02
    A coordinator fuses them
    It weighs the specialists in the light of market conditions, or abstains.
  3. 03
    The result is tested honestly
    No future information, walk-forward testing, results net of costs.
About the lab

Our mission is to build AI for finance whose claims survive realistic conditions, and to develop evaluation standards the field can adopt.

Forecasting edges in finance have a long history of disappearing once costs, data snooping and look-ahead bias are accounted for. We treat that as our starting point, and we value a well-tested negative result as much as a positive one.

/01

Pre-specified

Hypotheses and tests are fixed before results are seen.

/02

Out of sample

Walk-forward testing, with no information from the future.

/03

Net of costs

Economic results reported across realistic cost assumptions.

/04

Open about limits

We say what our data can and cannot detect.

Research

Four research programmes.

Each programme asks its own question and runs its own projects. Together they share one standard of evidence.

P1Active

Evidence and evaluation

Systematic reviews of AI in finance that measure how the field tests its claims.

Project: evidence review →
P2Active

Multi-agent forecasting

Specialist agents for fundamental, technical and sentiment evidence, and a coordinator that must beat a simple average net of costs.

Project page coming soon
P3In development

Language models and financial disclosure

What language models can reliably extract from company reports, filings and other financial text.

Project page coming soon
P4Planned

Open evaluation standards

A shared testing harness and reporting checklist that other researchers can use for AI forecasting claims.

Project page coming soon
Featured project · Programme P1

How rigorously is AI for stock prediction tested?

A systematic review of 775 journal studies, 2015 to 2026. Its finding: the field's methods are advancing faster than its reporting of how they are evaluated.

Read the project
775

journal studies reviewed

9 in 10

do not state how models were evaluated

< 3%

meet a full evaluation standard

2026

walk-forward reporting rises sharply

Publications

Papers and reports.

Team

Our team.

AAAmir Amel-Zadeh
Lab lead

Amir Amel-Zadeh

Associate Professor of Accounting, Saïd Business School

Amir is Associate Professor of Accounting at Saïd Business School and a Research Fellow of Green Templeton College, Oxford. His research covers financial and non-financial disclosure, accounting and capital regulation in banking, sustainable investment, and the use of natural language processing and machine learning on financial disclosures. He holds a PhD and MPhil in Finance from the University of Cambridge, where he was a University Lecturer at Judge Business School, and worked at Lehman Brothers before entering academia. He is a board member of the UK Endorsement Board.

RKReza Kalantarinejad
Researcher

Reza Kalantarinejad

Research Assistant, Saïd Business School

Reza is a Research Assistant at Saïd Business School, researching the role of AI in innovation and agentic systems. At the lab he builds the agents, the evaluation harness and the systematic reviews. He holds a PhD in Mechanical Engineering from Amirkabir University of Technology, with a computational thesis on nano-biosensor design, and has more than 30 publications, three books and six patents. He co-developed the Innovation Readiness Level (IRL) model and holds a Scale AI STEM Fellowship.

Join us. We welcome visiting researchers, students and collaborators.

Get in touch
News

Latest.

October 2026

AI4Finance Lab launches its website

An introduction to the lab's mission, research programmes and team.

October 2026

Systematic review working paper

Our review of 775 studies on AI-based equity prediction is in preparation for journal submission.

Work with us.

We welcome researchers, practitioners and data partners interested in rigorous evaluation of AI in finance.

[CONTACT EMAIL]