• High throughput Rust service work

• Owned and optimized the underlying storage layer of a RocksDB-backed geocoding service, focusing on performance tuning, operational stability, and resource utilization.

• Delivered a Rust sidecar service causing a 25% drop in p90 latency e.g. 168ms to 131ms and 15% to up to 40% drop in memory utilization with 10-15% throughput increases for our main geocoding service with decreases in CPU usage and overall latency.

• Collaborated on a sagemaker finetuning and training pipeline for our new BERT based ML models in Python.

• Released an interactive Tamagotchi-style game using Bevy for the WASM client, Stripe for the payments, and egui for the UI. Collaborated with a game designer in France: https://vpetmon.com

• Contributed to Bevy, associated Bevy 3rd party libraries, and egui.

• Managed and setup multiple NixOS servers on Linode with custom images. Implemented a HAProxy reverse proxy and MariaDB database with Borg backups. Developed and deployed a custom Rust axum payment and game server using nix flakes and deploy-rs. Utilized Netlify for the CDN to enhance content delivery speed.

• Collaborated on developing a machine teaching (optimal ml model creation) application using Rust and React, focusing on NLP, in a small, agile team at a seed-stage startup, resulting in significant improvements in application performance.

• Optimized Rust backend codebase, improving system performance by over 10% across all user flows by improving code generation and data structures in critical sections.

• Enhanced query language by building a regex extension for parts of speech tag matching, allowing for new types of search within the product.

• Designed and implemented an NER-based query suggestion system improving information extraction efficiency, leveraging N-grams and advanced machine learning techniques.

• Developed an active learning ML Python service leveraging WebSockets using a model created with TensorFlow via Keras, enhancing real-time data processing efficiency from an existing model.

• Engineered real-time distributed systems using Go, Elixir, Ruby, and Erlang, improving the marketing platform performance with ScyllaDB as the primary datastore.

• Developed an Elixir event processing service using Kafka handling over 100M events daily; contributed enhancements to popular Erlang/Elixir Kafka clients expanding client features; established observability, autoscaling, and monitoring to improve system reliability.

• Enhanced the company internal programming language for the product recommendation service through grammar, parser, lexer extensions and a distributed cache implementation, boosting system performance in collaboration with another developer.

• Addressed high latency issues in the recommendation service by implementing a new Geospatial database, reducing geospatial operation times to 10% of the original enhancing system functionality and improving user experience.