What I've Been Up To

Current roles

1. University of Cambridge · Oct 2025 to present

   Visiting Master's Researcher

   Analysing the effects of fine-tuning methodologies on AI bias guardrails in LLMs, in collaboration with Noah Broestl, under the supervision of Dr. Umang Bhatt.

   PhilosophyComputation RelatedMeasuring model bias

2. Responsible AI Institute · Jan 2026 to present

   Research Engineer in AI Policy and Governance

   Building the theoretical and technical foundations for AI governance mechanisms, such as a risk classification and policy generation framework for agentic AI systems. More on the RAI page.

   PolicyComputation RelatedARCTrustXROARDecentralised AI governance

Research projects

Newest first.

1. Master's dissertation · TRACE Lab, Cambridge · May 2026 to present

   Analysing the Effects of Fine-Tuning Methodologies on AI Bias Guardrails in LLMs

   Testing the effectiveness of LoRA and OFT on eroding pre-programmed AI bias guardrails in LLMs. Aiming to develop a practical recommendation and theorem that industry and academic stakeholders can use to determine how much fine-tuning data they need to eliminate bias guardrails in their specific use case. In collaboration with the TRACE Lab at the University of Cambridge.

   PhilosophyPolicyComputation RelatedMeasuring model bias

2. Imperial College London · ELEC70122: ML for Safety Critical Decision-Making · Jan 2026 to Mar 2026

   Developing Bias Identification and Mitigation Techniques for Clinical Prediction Models (CPMs)

   Devised bias identification methods that showed CPMs can learn to use missing data as a predictive signal and contribute to undesirable feedback loops in clinical settings. Introduced an uncertainty-triggered measurement intervention and a causal RLHF pipeline as bias mitigation methods for this use case.

   • Coursework paper (NeurIPS format, per the coursework brief)

   PhilosophyPolicyComputation RelatedMeasuring model biasFairness for underrepresented groups

3. Penn HCI Lab · Aug 2024 to Feb 2026

   Investigating the Presence of Bias and Potential Copyright Concerns in LLM Image Generation Capabilities

   Used quantitative and qualitative sociotechnical evaluation methods to investigate racial bias, gender bias, and potential copyright concerns in LLM-generated movie posters.

   • Poster and oral presentation, Penn NRCP 2025
   • Work-in-progress manuscript

   ComputationPsychology RelatedMeasuring model biasFairness for underrepresented groups

4. Penn Psychology Department · May 2023 to Sep 2023

   Utilising Correlational Analysis to Identify Traits of Successful Forecasters

   Applied correlational analysis to identify the traits and behaviours most associated with successful forecasters in a forecasting tournament.

   • Poster presentation, CURF Research Expo 2023

   Cognitive sciencePsychology RelatedHuman and machine cognition

5. Penn Medicine · May 2022 to May 2023

   Comparing Predictive Machine Learning Models' Capacity to Objectively Predict Dyspnea

   Developed predictive ML models that automatically and accurately estimated a patient’s breathing exertion levels, allowing doctors to monitor patients with respiratory illness without using physically invasive methods.

   • Poster presentation, BMES 2023 Annual Meeting

   NeuroscienceComputation

Skills

Awards

• University of Pennsylvania · Aug 2024 and Aug 2025

  Rhodes Scholarship Nominee

  Endorsed by UPenn to move forward in the Rhodes Scholarship process.

• Phi Beta Kappa · May 2025

  Phi Beta Kappa Inductee

  Top 8% of UPenn's graduating class of 2025.