Joshua YeatsPortfolio / 2026
Machine learning for the messy, human side of healthcare.
I'm a Data Scientist and Machine Learning Engineer at Dexcom, working across applied research and product development for time-series ML.
Interactive glucose simulation
Dexcom / 2025—PresentData Scientist + ML EngineerRemote / United Kingdom
Profile
- Current role
- Data Scientist / Machine Learning Engineer, Dexcom — 2025 to present.
- Practice
- Applied research and machine learning product development.
- Toolkit
- Python, PyTorch, JAX, TensorFlow, SQL, GCP and GPU-accelerated experimentation.
- Education
- BSc (Hons) Artificial Intelligence, University of Edinburgh, 2025.
- Location
- Remote / United Kingdom.
Selected Work
- 01 / 04Honours Dissertation
Interpretable Decision Support
Benchmarked program-synthesis rules, LLM-generated rules and Q-learning policies for Type 1 diabetes, with an emphasis on interpretability and time in range.
- Python
- SyGuS
- Q-learning
- LLMs
- 02 / 04iOS Application
Islet Diabetes
Privacy-focused iOS companion integrating CGM, heart-rate, activity and meal data, with personalised insight and medication-tracking workflows.
- Swift
- SwiftUI
- HealthKit
- FastAPI
- 03 / 04Research Code
LSTM_BGF
LSTM forecasting for noisy CGM data, supported by trend, rate-of-change and contextual feature engineering.
- Python
- PyTorch
- pandas
- LSTM
- 04 / 04Research Code
xLSTM_BGF
Extended LSTM architecture with attention and residual connections for more stable, personalised glucose prediction.
- Python
- PyTorch
- xLSTM
- Attention