Julien Cornebise

A fully integrated profile spanning the whole arc of AI: from fundamental methodological research (Bayesian machine learning and computational statistics, published in Nature and at NeurIPS and ICML), to deployment in high-stakes domains (medicine, human rights, industry, space), to the wider ecosystem around it (academic grant-giving, venture capital from seed to scale-up, and non-profits), with hands-on track record in each. One of the earliest researchers at DeepMind (2012–2016); my work contributed to raise significant early funding which then led to the acquisition by Google in 2014. A founder, builder and leader of teams, with a track record of hiring, mentoring and growing talent, and of setting research direction across research labs, start-ups and non-profits.

Combines deep technical rigour with a strong ethical compass and cross-disciplinary range, from mathematical foundations to engineering and to the societal implications of technology.

Independent AI Researcher & Consultant, Founder & CEO at WhyHow Ltd (2019 – Present)

Strategic and hands-on technical AI consulting to technology start-ups, NGOs and venture-capital firms (speech recognition, vaccine development, cancer detection, satellite imagery), including scientific due diligence for investors. Examples of engagements:

• R&D division of a Fortune-50 pharmaceutical company (2023–2025): Bayesian pharmacokinetics / pharmacodynamics; neural approximations of ODE systems; sequential Monte Carlo in JAX (BlackJAX, Diffrax).
• Joint venture between a Swiss start-up and a large human-rights archive (2023–2025): bioacoustics ML (YAMNet, AudioSet); event detection of specific conflict-zone activities.
• Cambridge-based, Series-B, best-in-class ASR (speech-recognition) start-up (2023–2024): speech and audio-LLM advisory; tokenisation, sample efficiency, SALMONN, Whisper.
• European Space Agency, QueryPlanet (2021–2023): deep-learning super-resolution of satellite imagery. Output: the WorldStrat dataset (NeurIPS 2022) and the HighRes-net open-source toolbox.
• Formulation and implementation of four big-data coefficients for crowdsourcing and robust dataset creation, with Klaus Krippendorff (2020–2021).
• Seed-stage biotech start-up (2020–2021): designed their first quantitative approaches (precision-oncology and genomics); interviewed and hired their early ML engineers.
• Investment due diligence:
• Recurring technical due diligence for several venture-capital firms and family foundations, including a large-scale sustainability growth fund; typically on Series-C-level start-ups.

Co-Founder & Chief Scientific Officer at ShiftLab (2021 – 2023)

• Co-founded and scaled an AI-driven product lab from 2 to 31 employees; managed a total budget of approximately £6 million.
• Set and owned the company’s scientific direction and technical backbone.
• Built the science team; led recruitment and technical interviewing.
• Executive leadership and strategy alongside the CEO (two co-founders).
• Three product verticals:
• Computer vision for industrial X-ray defect detection on polymer fabrics.
• Forecasting for container markets.
• Generative AI for end-to-end marketing campaigns.

Director of Research, Head of the London Office at Element AI (2018 – 2019)

Element AI (founded by Yoshua Bengio, 2018 Turing Award) Created and led Element AI’s London office and its “AI for Good” division; built the team from 1 to 12.

• Projects:
• Troll Patrol (with Amnesty International): analysis of online abuse against women; front-page Financial Times.
• Satellite-imagery super-resolution (European Space Agency): ESA-challenge finalist; led to HighRes-net and the WorldStrat dataset.
• Shaqodoon (Somali NGO): automatic speech recognition and classification in Somali; reached the MIT Solve final (top ~4% of applicants).
• Forensic Architecture: classification of tear-gas canisters used against civilians.
• Human Rights Watch: automatic analysis of video data and satellite video streams.
• Global Witness; World Bank (SDGs); UN World Food Programme; Oxfam.
• DECoVaC: design of experiments with controlled variability; reproducibility in few-shot learning.
• Partnerships & strategy:
• MILA; the Partnership on AI; AI Commons; the Montreal Data License.
• Media:
• Front-page Financial Times; Fortune; The Guardian; BBC; TechCrunch; Montreal Gazette; La Presse.

Senior Research Scientist, Frontiers Research Group; Head of the Health Research team, Senior Research Scientist at DeepMind Technologies Ltd then Google DeepMind (2012 – 2016)

• 2012–2015: Senior Research Scientist, Frontiers Research Group
• One of the five earliest researchers (employee #36). A range of early research, including the company’s first research patents, contributed to the early demos and to the significant fundraising that preceded the 2014 acquisition by Google.
• Heavily involved in building the team: helped design the technical screening interviews and conducted extensive interviewing.
• 2015–2016: Head of the Health Research team, Senior Research Scientist
• Co-founded and led the Health Research team, growing it from 1 to 10; set its research direction and technical backbone, matching DeepMind’s capabilities to clinicians’ needs under heavy privacy and security constraints. Work with the NHS / Moorfields (retinal diagnosis) and US Veterans Affairs (acute-kidney-injury prediction), widely covered in the press.

Independent Mathematical Consultant (2012 – 2012)

For worldwide pharmaceutical statistical and bio-statistical firms. Methodological and algorithmic expertise, data analysis, and software implementation, in Bayesian and more classical statistics.

Postdoctoral Research Associate at University College London, Department of Statistical Science (2011 – 2012)

Postdoctoral Fellow at SAMSI (Duke University, NC) and University of British Columbia (Vancouver) (2009 – 2010)

A two-year joint postdoctoral fellowship on Sequential Monte Carlo methods: first year at SAMSI (active participant in the SMC annual programme), second year at UBC under Pr. Arnaud Doucet.

Scientific & Technical Advisor at The Computational Democracy Project (Pol.is) (2021 – Present)

Long-running advisory to the open-source Pol.is platform for large-scale online deliberation.

• Research roadmap and strategy; reproducibility engineering; hands-on implementation.
• Research on the opportunities and risks of LLMs for scalable deliberation with Pol.is (2023).

Honorary Associate Professor (formerly Honorary Research Associate) at University College London, Department of Computer Science (2017 – Present)

• PhD examining: two early-assessment (upgrade / transfer) vivas.
• Temporarily co-supervised a PhD student (project later became FairGEN).
• Advise faculty on grant applications; mock-interviewed UCL UKRI Future-Leaders-Fellowship candidates.
• Member, ELLIS (European Lab for Learning & Intelligent Systems).

Scientific advisory boards & mentoring at ZOE; Synteny (computational-immunology ML) (2017 – Present)

• Scientific advisory boards: ZOE; Synteny (computational-immunology ML).
• Informal scientific advisor to numerous start-ups and research projects.
• Mentoring of early-stage founders and young executives, and of early-career researchers.

Pro Bono Scientific Advisor at Amnesty International (Amnesty Tech) (2016 – Present)

Ongoing AI and data-science expertise for human-rights investigations and advocacy.

• Decode Darfur: detection of destroyed villages from satellite imagery.
• Detected and mapped NYPD surveillance cameras through crowdsourcing; the resulting data was used in a lawsuit against the NYPD.
• Large scale data scraping and labelling.
• TikTok and teen mental health: analysis used to help initiate EU regulation.
• Russian-speech war-crimes project: analysis of intercepted communications.
• Designed the investigation into Denmark’s welfare algorithm and reviewed the report.

Ph.D. in Mathematical Sciences; Statistics and Applied Probability – University Paris 6 (2005 – 2009)

M.Sc. in Mathematics and their Applications, Speciality Statistics – University Paris 6 (2004 – 2005)

M.Eng. in Computer Science & Engineering Major – École Supérieure d’Informatique Électronique Automatique (ESIEA) (2000 – 2005)

French National Baccalauréat in Scientific Section (2000-07 – 2000-07)