Brian – Machine learning, Data Science, Big Data

A multi-strategy hedge fund management firm that now focuses on delivering a financial platform for investors.

• Developed and deployed machine learning models and large language models (LLMs) for financial report categorization, achieving 95% accuracy and improving stakeholder decision-making efficiency;
• Built a modular framework to process financial datasets, reducing preprocessing time by 25% and enabling quicker deployment iterations.

The world’s leading digital cross-device graph. It enables marketers to identify a brand customer or related household across multiple devices, unlocking critical use cases across programmatic targeting, media measurement, attribution, and personalization globally.

• Deployed Ray Serve to orchestrate online (real-time) and batch inferences, integrating with Flink streaming pipelines and Spark workflows. Achieved 99.9% uptime and reduced serving costs by 30% via autoscaling on Kubernetes.;
• Leveraged Ray Train for distributed training of GNNs and embedding models across 100-node clusters, reducing training time by 50% while maintaining model accuracy for identity resolution.;
• Engineered offline (Spark) and online (Flink) graph systems, with Ray Data preprocessing batch inputs and Ray Serve dynamically updating online user profiles, ensuring <100ms latency for ad-targeting use cases.;
• Unified training and serving with TFX pipelines, integrating Ray for seamless transitions between batch inference (daily user clusters) and real-time updates (event-driven triggers).;
• Built Spark workflows for petabyte-scale datasets, paired with Ray Cluster autoscaling to reduce resource waste by 25% during peak inference workloads.;

Data & AI platform solutions for various IBM external clients across diverse industries for ensuring the scalability of their data and machine learning models.

• Designed a low-latency nearest neighbor recommendation system achieving sub-100ms response times.;
• Integrated K-Means clustering for FAISS index optimization, enhancing system scalability for high-traffic workloads.;
• Designed Kafka-based streaming updates for FAISS partitions, enabling zero-downtime index refreshes.;
• Delivered an AI helpdesk portal (React, NodeJS, Watson Discovery) with NLP clustering, reducing query resolution time by 35%.;
• Deployed Airflow on Kubernetes and introduced Kedro for scalable terabyte-scale pipelines, improving team productivity by 20%.;

The team provided data modeling solutions to various external clients in multiple industries through IBM, addressing their specific business use cases.

• Built ML models (XGBoost, Random Forest) for oil well failure prediction, improving accuracy by 11% via custom metrics and NLP feature engineering.;
• Developed an NLP classifier using TensorFlow, achieving 92% accuracy, and built an Angular dashboard for legal teams as an MVP, securing a business deal with IBM valued at over $1 million.;
• Automated rule-based analytics on DB2, identifying 1,200+ high-risk cases for government audits.;

The revenue service of the Canadian federal government, and most provincial and territorial governments. The CRA collects taxes, administers tax law and policy, and delivers benefit programs and tax credits.

• Built an ML pipeline (Scrapy, Scikit-Learn) to flag non-compliant businesses, improving audit targeting efficiency by 30%, recovering over $1 million dollar back in taxes.;
• Implemented a web scraping solution using ScraPy, BeautifulSoup with features like rotating proxies, dynamic user-agents, and rate limiting to handle anti-scraping mechanisms and ensure reliable data extraction to support auditors.