Samir Qureshi — Machine Learning Engineer
London
Machine learning engineer with seven years putting models into production in healthcare and retail. Owns the referral triage model at a health technology company, where urgent cases now reach a clinician in a median of 40 minutes rather than six hours.
Experience
Machine Learning Engineer, Lumenstack Health · June 2023 – present
Clinical models team of six. Owns triage ranking, the feature store and the release pipeline behind both.
- Owns the triage model that orders 9,000 daily clinical referrals; urgent cases now reach a clinician in a median of 40 minutes rather than six hours.
- Built the offline and online parity checks that took training-serving skew from 4.1% of predictions to 0.3%.
- Cut inference cost 64% by distilling a 340M-parameter classifier into a 22M student with no measurable loss of recall at the operating threshold.
- Set up the shadow-deployment pipeline every release now passes through; it caught two models regressing on under-18 referrals before any patient saw them.
Machine Learning Engineer, Bramblewick Retail Group · March 2021 – May 2023
- Rebuilt size recommendation across a catalogue of 1.4 million items, cutting fit-related returns from 23% to 16% and saving £2.1m a year in handling.
- Replaced a nightly batch scorer with a streaming pipeline, so the first three clicks of a visit changed what a shopper was shown in the same session.
- Wrote the drift monitoring that flagged a broken colour taxonomy four days before it would have shown up in sales.
Data Scientist, Wrenfield Insight · September 2019 – February 2021
- Built the churn model for a subscription business; its top decile took a retention offer at 3.4 times the control rate.
- Automated the weekly modelling refresh, turning a two-day manual routine into 20 minutes and freeing an analyst for experiment design.
Projects
- Driftgauge — An open-source feature-drift monitor for scikit-learn and PyTorch serving pipelines.
Education
- MEng Computer Science · University of Manchester · September 2015 – June 2019
Skills
- Modelling: PyTorch, scikit-learn, XGBoost, Distillation, Calibration
- Production: Kubernetes, MLflow, Feature stores, Ray Serve, Shadow deployment
- Data: Python, SQL, Spark, Airflow, Snowflake
- Practice: Drift monitoring, Experiment design, Clinical safety review, Code review