Apr 2025 – Jul 2025 · 3 mos

• Part of a curated community of exceptional future founders, working to build globally impactful companies together.
• Entrepreneur First is the world’s leading talent investor, backed by top founders including Patrick and John Collison, Reid Hoffman, Tom Blomfield, Sara Clemens, Nat Friedman, and Matt Mullenweg. EF brings together people with extraordinary futures to develop their most ambitious ideas and raise money from top global investors.

Oct 2023 – Apr 2025 · 1 yr 6 mos · London, England, United Kingdom

• Created and maintained a template for RAG-based projects using Azure, OpenAI, and LangChain, saving an estimated week of development time for new projects.
• Wrote an evaluation suite for indexes and benchmarked Annoy and Redis on several metrics such as speed and accuracy. Reported the results to the team for analysis.
• Created an efficient flat index from scratch with various customizability using NumPy.
• Identified a lack of documentation and standards around various team practices and proposed solutions, increasing team efficiency through increased code quality.
• Researched bid writing and analysis products in-depth and presented findings to senior stakeholders. Designed qualitative experiments to evaluate their effectiveness.
• Learning Course Generator
• Designed a bespoke solution for an existing learning course creation process that generates a course script from a high-level document using Azure, OpenAI, LangChain, and React, reducing script creation time from 6 to 2 days.
• Designed and implemented a dynamic framework that enabled 26 different course pages to be generated, formatted, and exported as a Word document in 2 weeks in Python.
• Created a modular prompting engine that dynamically builds prompts via parameters such as method (e.g., Chain-of-Thought), context, styling guidelines, etc. Increased output robustness by implementing JSON extraction methods using a combination of LLMs and RegEx.
• Implemented G-Eval, a state-of-the-art LLM evaluation technique, and used it to evaluate hundreds of generated course pages, increasing evaluation reliability and content quality.
• Bid Writing Assistant
• Created a full stack bid writing product that can generate sophisticated bid plans and responses in less than a minute using Azure, OpenAI, LangChain, Next.js, and React.

Sep 2022 – Oct 2023 · 1 yr 1 mo · London, England, United Kingdom

• Synthetic Data
• Developed a novel algorithm for tabular data imputation using PyTorch and scikit-learn that significantly outperformed previous methods used at the bank (+0.4-1.0 R-squared).
• Implemented and documented a comprehensive synthetic data evaluation suite using PyTorch and scikit-learn, and Pandas. Contributed to a novel evaluation method currently being patented.
• Used NVIDIA’s NeMo and the SDV frameworks to implement synthetic data generators such as Megatron and CTGAN in Python.
• Wrote unit and integration tests using unittest for the imputation algorithm and evaluation suite (90%+ coverage).
• Engaged with stakeholders and the team throughout and presented the evaluation suite, imputation algorithm, and results to Global Risk Analytics (organisation of 4000+ people).
• Code Metrics & Model Automation Dashboard
• Enhanced code quality across the organisation by creating a dashboard that displays code metrics such as repository activity, number of bugs, duplicated blocks, and coverage using React and Oracle DB. Wrote unit and integration tests using Jest (90%+ coverage).
• Created a dashboard from scratch that automates model calibration, execution, validation, testing, and reporting in React, Next.js, and Oracle DB.

Jan 2022 – Aug 2022 · 7 mos · Edinburgh, Scotland, United Kingdom

• Tutored fourth-year students natural language fundamentals and assisted them in the lab with Python and NLTK for the module Foundations of Natural Language Processing.
• Tutored first-year students mathematics: Calculus, Linear Algebra, Probability, etc.

Sep 2020 – Dec 2020 · 3 mos · Liverpool, England, United Kingdom

• Investigated hand-eye calibration techniques to minimise the error associated with transforming from world coordinates to homogeneous coordinates and vice versa, especially in error-prone environments such as the pick and place of small items.
• Became familiar with image formation, planar homography, and hand-eye calibration methods, and implemented some of the methods with Python and OpenCV.

Sep 2015 – Mar 2016 · 6 mos · Exeter, England, United Kingdom

• Debugged and tested custom in-house PHP projects and created custom plugins in PHP for platforms such as WordPress.
• Provided support to clients online and over the phone needed excellent communication and understanding of customer expectations.
• Tutored an apprentice for 3 months.