As a data science consultant helping AWS customers deliver business value with AI/ML:Structured data and time series:- Built a probabilistic predictor for football return yards using convolutional networks and rare event modeling - Led the implementation of a bond trading recommender using graph machine learning (Knowledge Graphs, Graph Neural Networks)- Developed t-link, a simple and competitive method for Link Prediction in Temporal Graphs - Built a fraud detector for a leading prepaid debit card company (SQL, Gradient-Boosting)- Designed and implemented an optimization algorithm for semiconductor testing (feature selection, discrete optimization) Natural Language and text:- Built a document chatbot demo based on generative AI (ChatGPT, Amazon Bedrock, HuggingFace, LangChain)- Built an adverse event predictor using medical claims data (CMSAI competition 2020, PyTorch)- Designed and prototyped a solution for medical code relabeling using clustering and representation learning- Surveyed and prototyped deep learning techniques for clinical data including ClinicalBERTComputer Vision:- Led the implementation of a video recognition pipeline for soccer games (I3D, SlowFast)- Provided statistical guidance to a visual anomaly detection project (unsupervised computer vision, PaDiM)

As the head of data science within the digital transformation team:- Cut capital cost by implementing machine learning forecasting on business loans- Improved trading backtest profits by creating and tuning a machine learning prototype - Built a product and client recommender for the debt market- Co-created the AI strategy and capability for the wholesale banking divisionTechnology: Python, Scikit-learn, Tensorflow/Keras, R, Flask, SQL, Hadoop, Hive, Spark, GPU

As the leader of a quantitative risk software team:- Cut the annual running cost while increasing quality ratings by designing, prototyping and leading the creation of the next risk computing platform for trading desks worldwide - Built and led a team of up to 18 software engineers, managed cross-functional project teams of up to 30 engineers covering quantitative, software and operationsTechnology: C#, SQL, Cassandra NoSQL, web services, private and public cloud computing

- Transformed a legacy computing system into a service oriented scalable distributed platform (runtime divided by 5) by rewriting the distribution component in C++ and Java - Cut the time to market for risk analytics from 2 weeks to 1 day by creating a scripting environment

Led the study and prototyping of a structural re-design for a large distributed computing platform (C++, Java)

Created interactive TV apps for our main product (Linux, C, Javascript, HTML)Netgem was a high growth tech startup that IPOed in 2001

Created prototypes to demonstrate internet technology equipment to telecom operators (Unix, Cisco OS)