Daniel Gaskins

Applied Machine Learning & Research Engineer

hello@danielgaskins.com danielgaskins.com United States · Open to relocation

Summary

Applied machine learning and research engineer with founder-level ownership across model development, evaluation, data workflows, deployment, and user-facing products. Physics-trained experimentalist who turns ambiguous product questions into measurable ML problems and carries promising models into dependable software.

Technical Skills

Machine Learning: Model training and evaluation, experiment design, error analysis, performance optimization, NLP, computer vision, gradient-boosted trees

Frameworks: PyTorch, TensorFlow, scikit-learn, LightGBM

Software & Systems: Python, C++, JavaScript, TypeScript, Scala, Java, React, data workflows, inference, deployment, production engineering

Professional Experience

Perfsy — Founder & Applied Machine Learning Engineer

2021–Present

  • Build ML-driven performance optimization systems across model development, data workflows, backend services, and product integration.
  • Design and run experiments, evaluate model behavior, inspect failure cases, and turn findings into successive model and product iterations.
  • Own technical decisions from early research and prototyping through production engineering, deployment, and user feedback.

Namelor — Founder & Product Engineer

2019–2021

  • Built and operated a software product end to end, including product requirements, interface development, backend architecture, and deployment.
  • Translated ambiguous user needs into technical decisions and shipped iterative improvements across the full application stack.

Casabauhaus — Founder & AI Research Engineer

2017–2019

  • Prototyped and tested AI-assisted design tools intended to support architectural exploration without flattening the human creative process.
  • Worked across research engineering, model behavior, interaction design, and product development to evaluate technically and creatively useful approaches.

Selected Open-Source Project

lgbm-to-code — ML Deployment Tooling

pypi.org/project/lgbm-to-code

  • Developed a code generator that converts trained LightGBM models into plain Python, C++, or JavaScript conditionals.
  • Enabled portable inference without a runtime dependency on LightGBM.

Education

University of California, San Diego

B.S., Condensed Matter Physics