/deveshbatra/
I work on applied AI systems: LLMs in finance, responsible deployment, model behaviour, uncertainty, and decision-making under risk.
Currently: NatWest AI Research · Oxford DPhil
2025 :: Workshop paper :: NeurIPS 2025 Workshop on Generative AI in Finance
A production-focused paper on LLM-assisted drafting for compliance-critical financial-services letters.
2025 :: SSRN preprint :: SSRN
A survey of LLM agent applications in finance and banking, organised around simulation, acting, analysis, and advising.
2025 :: Conference paper :: Findings of ACL 2025
Studies how LLM personality traits relate to risk propensity using Cumulative Prospect Theory and Big Five framing.
2025 :: arXiv preprint :: arXiv
Evaluates latent nationality bias in LLMs when explicit demographic labels are replaced by culturally indicative names.
2023 :: arXiv preprint :: arXiv
Introduces Longitudinal Predictive Conformal Inference for finite-width prediction intervals with longitudinal and cross-sectional coverage goals.
03/05/2026
Why keep a personal research homepage in the age of LinkedIn and Scholar.
$ focus --now
Research code :: archive
Reference code for conformal prediction over longitudinal data.
Casual webapps :: idea
A small webapp for creating accessible match lists, rotations, or doubles pairings for a badminton club.
AI / data demos :: idea
A public notebook-style explainer for thinking about agent usefulness, risk, and deployability in finance.
I like small, legible systems.
Research should be understandable. Software should be useful. AI systems should be measured where they are actually used. And side projects should be allowed to be weird.