As an early employee, I have worn many hats across scientific, ML, and software engineering disciplines, as the company scaled from 20 to >200 people and > $1B valuation. Sample projects I’m involved with:* Led the ML-guided reconnaissance program to support our flagship nickel exploration in Northern Quebec.* Developed ML models on multispectral satellite imagery for outcrop identification and lithology classification, and rapidly iterated the models with ground-truthing from our reconnaissance team, leading to discovery of previously unmapped high-grade mineral occurrences.* Architectured and implemented the machine learning pipeline for large-scale remote sensing ML end-to-end, to support continental scale (> 1M km2) prospecting of Ni, Cu and Li, and reduced the ML model turnaround time from weeks to hours.* Developed ML-informed geological mapping with unsupervised clustering on airborne geophysics and remote sensing rasters; customized loss metrics for evaluation of clustering models against independent, sparse field observations. This enabled bootstrapping of large-area greenfield exploration activities with little ground truth. * Designed database schema and scalable data pipeline for processing and ingesting airborne geophysical surveys; system is used by the entire DS team for ingesting > 200 airborne geophysics surveys across the globe.

* Lead data scientist for developing and deploying machine learning models for payment fraud defense of our entire credit card portfolios in US and Canada. * Built reusable, end-to-end model development pipelines, including infrastructure provisioning on AWS-EMR, data pull and feature engineering code in PySpark and SQL, supervised machine learning models in H2O, and model monitoring stack in Python, InfluxDB, and Grafana.* Coordinated model deployment on AWS-based platform and conducted model validation in prod; captured bugs introduced during the feature function rewrite in Java, thus allowing on-schedule model deployment.* Intimately involved in data scientist recruiting processes for the entire enterprise, serving as one of a handful of interviewers for majority of on-site DS interviews, and provide feedback to shape our recruiting practices.

* Built prototype tools to consume customer digital interaction event streams on Kafka, and explored NLP / sequence models to generate insights to power personalized customer experiences over digital channels.* Interim product manager of in-house clickstream analytics platform that leverages Kafka and Snowplow. Coalesced efforts for monitoring and analysis, and coordinated user transition from legacy platform.

* Analyzed TBs of credit card transactions to identify characteristics and trends of block-level neighborhoods in selected US cities. Developed geospatial data pipelines in Python (fiona, rtree, shapely) and postgres / PostGIS, customer segmentation models in Python, and geospatial data-viz web app in R-shiny / leaflet. * Product owner & team lead for internal platform to automate workflows for business metrics monitoring and dashboards using Python, InfluxDB, and Grafana. Mentored 50+ analysts in 30+ teams, and implemented self-service instruction to scale adoption. (More details available on my PyCon 2019 talk)

* Developed statistical analysis and data visualization tool in R-shiny, reduced routine analysis time by 95%.* Built reusable data pipelines on manufacturing line data to connect multiple processing and testing steps, and developed tree-based models to provide insight on process control capabilities and improve yield.

* Performed electrical and optical modeling to predict and improve thin film GaAs solar cell performance.* Developed fabrication processes to improve solar cell efficiency, leading to 2 world records and 3 patents.