Sam Armstrong

Interesting Things I've Worked On

PDF Word Count

Created an extension for Visual Studio Code and Open VSX compatible editors (such as Cursor), which displays the word count of PDF documents in the status bar, along with more detailed statistics about a PDF (file size, page count, character count, etc.) in the status bar tooltip. The extension is published and available on the VS Code Marketplace and Open VSX Registry (see links below).
Skills
- TypeScript
- Azure DevOps
- CI/CD
- Open Source
- VS Code
- Open VSX
CozyCasts

Built CozyCasts, an AI-generated podcasts platform to allow the creation of custom podcasts for learning and entertainment, with features like voice customisation, automatic artwork generation, and saving podcasts to a library for future listening. The backend was created in Python and FastAPI, deployed on Google Cloud Platform, making use of the Gemini and Claude APIs. The frontend was created in React and TypeScript, deployed on Vercel.
- CozyCasts Website
Skills
- Python
- System Design
- GCP
- Agentic AI
- AI Deployment
- React/TypeScript
UFC Almanac

A collection of UFC datasets of fighters and fights scraped from the web and regularly updated via CI/CD pipelines in GitHub Actions. Also includes data transformation pipelines to convert the raw data into structured training datasets for machine learning, and deep learning models/training pipelines for predicting fight outcomes.

The primary model used is a transformer model that, for each fighter in an upcoming fight, encodes the sequence of statistics from their previous fights into a latent space, then uses a classification head to predict the outcome of the fight.
- GitHub
Skills
- Python
- PyTorch
- Neural Networks
- ML Pipelines
- Data Engineering
- Web Scraping

Writing

Research Proposal: Investigating Integrations of the Flang Parser with OP2

Sam Armstrong

2026
An Agentic Approach to a First Price Auction with Complementary Goods

Sam Armstrong

2026
Parallelising a Computational Fluid Dynamics Solver using Distributed Memory Parallelisation

Sam Armstrong

2026
A Survey of Language Models for Source-level Code Optimisation

Sam Armstrong

2025
3D Pose Estimation using mmWave Radar Data

Sam Armstrong

2025
Reconsidering an Agent-Based Model of Food Web Evolution

Sam Armstrong and Emma Norling

2022

Education

University of Warwick
2025-2026

MSc Computer Science

Currently studying for a Master's degree in Computer Science at the University of Warwick, focusing on advanced topics in high performance computing, agentic artificial intelligence, computer security, networks, and computer science theory. All standard modules (consisting of both exams and coursework, representing 2/3 of the course) have been completed with an average grade of 79% (strong distinction), with the thesis project currently in progress.

Notable modules
- Sensor Networks and Data Communications
- Agent Based Systems
- Machine Learning Applications
- High Performance Computing
- Computer Security
- Foundations of Computer Science
- HPC Thesis
University of Sheffield
2018-2021

BSc Artificial Intelligence and Computer Science

Grade: First Class Honours

Studied for a Bachelor's degree in Artificial Intelligence and Computer Science at the University of Sheffield, achieving First Class Honours with a final year grade average of 79.5%. The course focused on the theoretical grounding and applications of deep learning, software engineering, system design, and agent-based modelling.

Notable modules
- Agent-Based Modelling Dissertation
- Adaptive Intelligence
- Text Processing
- Speech Processing
- Bioinspired Computing
- Software Reengineering
- System Design and Security
- Functional Programming

Experience

Ivy
2024-2025

Co-Founder

Co-founded Ivy as a spin-off from Unify AI, after it pivoted into creating AI teammates for task management and delivery. We saw an opportunity to convert Ivy into a source-to-source transpiler for models and code between frameworks, using the multi-backend machine learning framework as a backbone and intermediate representation for this process. We successfully achieved a strong baseline of PyTorch -> {JAX, TensorFlow, NumPy} transpilations, and integrated Ivy into Kornia to allow their PyTorch-based computer vision library to be natively used with JAX, TensorFlow, and NumPy.

After interviewing with Y Combinator and talking to potential customers, we failed to prove commercial viability, so we brought Ivy to a close as a venture and open-sourced the project. It is now available on GitHub and PyPI, and I continue to maintain it in my spare time.

Skills
- Python
- TypeScript/React
- GCP
- Deep Learning Frameworks
- AST Manipulation
Unify AI (YC W23)
2023-2024

Machine Learning Software Engineer

Worked on developing the initial form of Ivy - a multi-backend machine learning framework. Primarily focused on creating a graph compiler capable of extracting efficient computational graphs from arbitrary Python code using function tracing, with features like subgraph tracing (to allow dynamic control flow to be captured) and automatic graph optimisation.

Some of my key responsibilities included interviewing candidates for both internships and full-time roles; attending events representing the company, such as a GCP event at Google in King's Cross and an AWS Startup event in Farringdon; meeting with investors and accelerators, such as Cherry Ventures and Intel Ignite; and reviewing open-source pull requests to Ivy.

Skills
- Python
- Git
- CI/CD
- Deep Learning Infrastructure
- Large Codebases
- Agile Development
Intradiem
2022-2023

Machine Learning Engineer

Developed machine learning data pipelines for converting raw data collected in a data lake into structured time-series formats to be used for training by the data science team and for inference in production. These were primarily created as Java microservices using Spring Boot and Temporal with Kibana for monitoring and observability, with some services written in Python using Flask and SQLAlchemy, and deployed to AWS EC2 using Docker containers and Kubernetes.

Skills
- Java
- AWS
- Python
- Docker
- Kubernetes
- SQLAlchemy
- Flask
- Microservices

Miscellaneous

PDF Word Count

CozyCasts

UFC Almanac

Research Proposal: Investigating Integrations of the Flang Parser with OP2

An Agentic Approach to a First Price Auction with Complementary Goods

Parallelising a Computational Fluid Dynamics Solver using Distributed Memory Parallelisation

A Survey of Language Models for Source-level Code Optimisation

3D Pose Estimation using mmWave Radar Data

Reconsidering an Agent-Based Model of Food Web Evolution

University of Warwick

University of Sheffield

Ivy

Unify AI (YC W23)

Intradiem