Summer 2025

• Developed a Rust based twirp search client, exposing their pre-existing search service progressively to the Rust SDK, Python SDK, and an internal CLI tool.
• Reduced search index size by 10x and eliminated long-standing race conditions on AWS S3, maintaining 100% data retrieval accuracy.
• Independently managed and delivered filtering in the search service based on user feedback, integrated support in both the internal CLI tool and React UI and presented the solution at company-wide weekly demos.
• Researched potential future full-text search solutions (OpenSearch, Manticore, LLM-based vector search) that allowed for sharding and inbuilt analytics and setup prototypes to demonstrate functionality and trade-offs.

January 2025

• Used random forests to predict lift movement between floors.
• Fused barometer and accelerometer data to find lift height.
• Braved High Wycombe to collect data from lifts.

Summer 2024

• Applied machine learning with an accelerometer to calculate lift kinematics.
• Conducted research, developed, and documented a paper titled 'Using Multivariate Polynomial Linear Regression to Model Asymmetric Lift Kinematics'.
• Released an app in Android Studio with Java to collect accelerometer lift data and analyse for customers.