Responsible for building AI/ML infrastructure and services to accelerate research, and translate breakthroughs into products.(Sep 2022- 2023) Led an organization creating a cohesive, interoperable, and composable suite of 1P ML products that seamlessly come together to help Googlers across all PAs (e.g., Research, Google DeepMind, YouTube, Ads, Geo) in their end-to-end ML journeys. Defined well-lit-paths (WLP) for ML at Google, and driving the WLP program at Core ML to accelerate developer velocity. Made strategic investments to support LM/GenAI needs and products are on the critical path for LM workflows at Google.(March 2023 - present) Created and established a new Trust/Responsible AI Infrastructure organization focused on the set-up and management of Trust for LMs/GenAI to ensure Safe AI launches. Developing products and tools for AI Governance with ML Lineage tracking and approvals workflows; used by all LM initiatives at Google. Developing infrastructure for GenAI Safety Classifiers to ensure generated content adheres to content safety policies. These classifiers are used to detect harmful, policy-violating user inputs or system outputs. (Sep 2023 - present)Responsible for the Core ML Training ecosystem infrastructure with products for ML experimentation, analysis, evaluations, and debugging, production training pipelines, self-contained accelerator workloads. Investing in strategic initiatives to improve ML performance and developer velocity.

Eigengene develops unique high-dimensional multi-tensor machine learning algorithms that connect a tumor’s whole genome with the patient’s survival and response to treatment. Our algorithms discover connections that are clinically applicable to the population at large from the whole genomes of cohorts as small as 50–100 patients. These accurate and precise connections outperform the best other clinical indicators, where they exist. All other methods miss them. For example, our retrospective clinical trial validated a genome-wide pattern in tumors from brain cancer patients as the best predictor of life expectancy and response to standard of care. We discovered this, and patterns in lung, nerve, ovarian, and uterine tumors, in public data.Eigengene’s algorithms and predictors will help patients and oncologists manage cancer and disease far more effectively than what is currently possible. For example, they can help avoid unnecessary surgery in response to pseudoprogression in brain cancer or decide against platinum-based chemotherapy in lung cancer. They can help clinical trials succeed by predicting which patients are likely to respond to a treatment. They can also identify new drug targets and combinations of targets that are correlated with survival.

Leads the Amazon ML Solutions Lab, a global team of scientists, engineers, and product managers who help AWS customers identify and implement their most important machine learning opportunities.

Leader on the Amazon Machine Learning Solutions Lab, whose mission is to work with companies and organizations from all industries to solve their most pressing business needs using machine learning.

Teach ML to business leaders and work to inspire the next generation of leaders.

Spearheaded the establishment of artificial intelligence capabilities at Genentech's Early Clinical Development team and Genentech Research and Early Development (gRED). Defined and led pilots designed to demonstrate the value of machine learning and deep learning to drug discovery and development. Projects supported clinical science and clinical operations and comprised of integrating diverse types of data, such as, clinical, biomarker, genomic, and imaging, and using artificial intelligence to answer key clinical and research questions, for prospective and retrospective clinical trials.

-Led team that implemented scalable machine learning algorithms & built models for content personalization and recommender systems for millions of customers as part of Chase Digital Intelligence. -Supported multiple LOBs and co-ordinated work across verticals of product, technology, marketing, and testing to deploy models to production. Model-based recommendations for Chase Ultimate Rewards marketing dramatically lifted redemptions while providing satisfying redemption experience to customers and adding to the bottom-line. -Technologies: Scala, Spark (MLlib), Hadoop, Hive

-Work with business to define, architect, build & deploy social media analytics solutions that measure & monitor impact of social media on financial markets. -Responsible for Bloomberg Social Velocity (BSV) that alerts clients of spikes in social activity & market sentiment about companies. Press:1. http://gigaom.com/2014/02/28/traders-turn-to-twitter-for-market-news-now-they-can-measure-mood-too/2. http://www.tradersdna.com/news/bloomberg-introduces-twitter-sentiment-analysis-tools/3. http://www.bloomberg.com/company/announcements/trending-on-twitter-social-sentiment-analytics/-Train sales & application specialists on products, mentor new hires, & provide machine learning consultation across firm.-Support recruitment initiatives with classroom & tech talks at universities.-Technologies: C, C++

Responsible for designing, building and enhancing several aspects of Bloomberg's risk analytics platform such as customized stress testing, risk scenarios generation and shock propagation, Monte Carlo simulation of risk scenarios for VaR computation, stress matrix valuation for non-linear securities, and risk factor models.

Developed generalizations of matrix and tensor computations: defined new tensor decomposition, higher-order generalized singular value decomposition (HO GSVD), that compares many large data sets and identifies similar/dissimilar patterns among them. HO GSVD extends GSVD (defined in 1976 for N=2 data sets) to N >= 2 data sets. Applied these to create models to analyze, compare and integrate different types of large-scale molecular biological data, such as DNA microarray data.

Upgraded product used to manage cable networks. Project outcome: 80% decrease in production cost + 60% increase in product sales.