abstract

Machine learning engineer with 7+ years taking models from research to production in NLP, recommendation and forecasting. Combines solid statistics with MLOps discipline and has published at applied-ML venues.

results

+18%CTR uplift
91%LLM routing accuracy
−23%stock-outs
12Musers served

experiment log

run-0042 · Apr 2022 - Presentbest model

Senior ML Engineer

Quantia Analytics
  • Shipped a two-tower retrieval and ranking system that lifted click-through rate by 18% in A/B tests.
  • Fine-tuned an LLM for support triage with 91% routing accuracy, saving 14 FTE of manual sorting.
  • Built a feature store and model registry, cutting time from experiment to production from 6 weeks to 5 days.
run-0031 · Jun 2019 - Mar 2022v2

Data Scientist

Meridian Insights
  • Built gradient-boosted demand forecasts for 40k SKUs, reducing stock-outs by 23%.
  • Designed an experimentation platform with CUPED variance reduction used by 9 product teams.
run-0017 · Jul 2017 - May 2019baseline

Data Analyst

Northwind Research
  • Automated survey analytics in Python and SQL, cutting report turnaround from 5 days to 1.
run-0001 · 2015 - 2017seed

M.Sc., Statistics & Machine Learning

Sample Institute of Science

models & papers

Real-time Recommendation Engine

Two-tower retrieval with online re-ranking serving 3k requests/s at a p99 latency of 40 ms.

PyTorch · Faiss · Kafka
LLM Support Triage

Fine-tuned a 7B model with guardrails, reaching 91% routing accuracy and running on-prem.

Transformers · LoRA · FastAPI