i build AI that does the boring stuff.

Self-Employed

• Help non-technical organizations bring AI into their workflows with bespoke software, automations and data infrastructure that get rid of mundane, repetitive work.
• For BioShift Life Sciences, built AI infrastructure from the ground up, including a retrieval-augmented research assistant that helps the team access and synthesize research across domains.
• Designed automations for discovering and digesting academic literature, cutting time spent locating and reading papers by 70%.

Sprout Social

• Reduced infrastructure costs by 45% ($80K annually) across 10+ production BERT models with FP16 mixed-precision inference, replica optimization (66 → 36), and quantization and distillation experiments, without sacrificing model quality.
• Improved model performance by +0.25 F1 and 2× precision through feature engineering, experimentation, A/B testing and post-deployment monitoring.
• Led evaluation and prototyping of LLM systems, designing benchmarks for quality, latency, bias, hallucination risk and cost to inform production decisions.
• Owned end-to-end Bluesky and Threads support for a production ML feature, extending ingestion pipelines and running Spark backfills for 6,000+ profiles.
• Built monitoring and observability for ML services with Airflow, StatsD and Datadog.

American Express

• Improved recommendation precision by 45% by developing and evaluating deep learning models for personalized product targeting.
• Engineered a two-tower recommendation model using user behavior and demographics to influence product strategy.
• Built scalable feature engineering pipelines, cutting data processing time by 30% and improving training efficiency by 20%.