May 2022 – Nov 2025 · 3 yrs 6 mos · Mountain View, California, United States

• 🚀 Democratizing ML on relational data
• At Kumo, our mission is to empower users with a seamless experience in constructing state-of-the-art Graph Neural Network (GNN) models on relational data. I am deeply passionate about leading innovation efforts to craft a world-class end-to-end product experience. Our customers can effortlessly import data, construct intricate graphs, and build high-performance predictive models using our cutting-edge Predictive Query language. The beauty lies in simplicity—just three lines of a query can train a GNN model on terabytes of data.
• 🔍 Opportunities Await! Join our Innovative Team
• We're on the lookout for forward-thinking individuals who share our passion for transforming the landscape of AI. If you're ready to make an impact, explore exciting career opportunities with us, apply here: https://kumo.ai/about/careers
• 🌐 Experience the Future of Data Analytics: Request a Demo
• Curious to see our product in action? Request a demo and discover how Kumo is revolutionizing the way data is harnessed for actionable insights: https://kumo.ai/apply

Jan 2021 – Present · 5 yrs 4 mos

Apr 2019 – May 2022 · 3 yrs 1 mo · San Francisco Bay Area

• In my role as Lead at Michelangelo, Uber's cutting-edge ML platform, I've played a pivotal role in driving innovation and enhancing the platform's capabilities.
• Embeddings Platform Transformation:
• Conceptualized and implemented an advanced embeddings platform for Michelangelo, showcasing a blend of innovation and practicality.
• Led the migration of hundreds of city-specific models for Uber Eats home feed ranking to a unified global Two Tower model, demonstrating a keen understanding of scalability and efficiency.
• Introduced various customizations and applied existing research, resulting in a substantial enhancement of overall performance despite the transition to a more generalized model.
• Spearheaded Uber's first productionization of a Two Tower model, a milestone documented in detail in the blog post: Innovative Recommendation Applications Using Two Tower Embeddings (linked below).
• Trust Region-based Bayesian Optimization (Turbo) productization:
• Successfully productionized and implemented the Trust Region-based Bayesian Optimization, as outlined in the Turbo paper (linked below).
• Conducted comprehensive benchmarks against other state-of-the-art Hyperparameter tuning models, establishing Turbo's effectiveness in enhancing model performance.
• This initiative had a significant horizontal impact, improving the efficiency of numerous mission-critical models across Uber.
• Thriving at the intersection of innovation and practical application, I am passionate about driving advancements in machine learning and contributing to the evolution of cutting-edge technologies. Let's connect to discuss how my expertise can elevate your team's capabilities.

Oct 2018 – Mar 2019 · 5 mos

• I was an engineer in the career relevance team where I was responsible for building machine learning models and tools and infrastructure to support career relevance.

Mar 2015 – Sep 2018 · 3 yrs 6 mos

• I was tech lead and architect for Hiring Platform, an ATS offering by LinkedIn, where I designed several mission critical projects in Hiring Platform like Rules engine, Notifications platform, Jobs Targeting and Sourcing channel integration. In addition, I was also one of the 10 Data Model Review Committee leads at LinkedIn who review and approve API changes across the company and help set tech strategy for APIs at LinkedIn.

Feb 2014 – Mar 2015 · 1 yr 1 mo

• I was Tech lead in the Companies Pages infra team. Most notably, I led Companies Pages auto creation project where the number of Company pages went up from 3.6M organic pages to over 18M. Auto creation pipeline fused company data from over 60 different sources into a single record after clustering the data using a Random Forest model. The clustered record would then be fused and converted to LinkedIn's entity taxonomy. This required several key innovations resulting in 4 approved patents.

Oct 2012 – Feb 2014 · 1 yr 4 mos · Mountain View, CA

• Worked in Google Wallet (now Google Pay). My most notable contribution was Google Wallet's authentication integration with Google's Single Sign on GAIA infrastructure.

Apr 2012 – Sep 2012 · 5 mos · Hyderabad Area, India

• Tech Lead for Transportation Invoice Processing System where I lead from scratch, development of a large scale system to automatically process transportation invoices for Amazon.

Aug 2010 – Apr 2012 · 1 yr 8 mos · Hyderabad Area, India

Jul 2009 – Jul 2010 · 1 yr · Mumbai Area, India

• Worked on an XMPP based chat server.

Jan 2009 – Jun 2009 · 5 mos · Bangalore

• Developed a JavaScript command line verification tool using JSLint and added output compatibility with JUnit.(php, JavaScript, JSLint)