About

Life

I have been living in San Francisco since 2018, reading fantasy fiction, long-distance cycling, taking photos, and learning Italian. My longest single day solo ride is 130 miles.

I pursued my Master’s in Computer Science at UCSD and Bachelor’s in Computer Science and Engineering at RVCE. I grew up in Udupi, a coastal quaint town in Karnataka, India, and Bangalore, the silicon valley of India.

Industry

Infinitus Systems - Staff Machine Learning Engineer

August 2022 - Present

AI Auto-Review

Currently leading the Auto-Review team, and we scaled from a single customer to 30+ customers across different tasks (benefits verification, prior authorization, pharmacy benefits management), reducing the need for human review to only complex cases. (technical report)

I enhanced our ML systems for scale with robust LLM + XGBoost models along with standardized training pipelines, optimized feature extraction (reducing training duration by ~75%), comprehensive observability, and alerting. I’ve actively mentored team members and championed psychological safety and inclusion.

Adverse Event Detection

Developed SAGE (Safety Adverse-event Guidance Engine) - a multi-modal (text+audio) NLP system to detect adverse events on healthcare voice agents with 99% recall and 0.3% false positive rate (in production). (blog)

Voice Call Automation

I have been involved in building the voice call-based conversational AI system, achieving 90% IVR, Hold and Authentication call phase automation with no human intervention. This involved building a scalable ML platform that supports rapid experimentation, while ensuring low latency and high reliability. I led the development of shadow mode testing for non-disruptive model deployment in production, along with scalable model service design. I incorporated advanced NLP solutions using Gemini prompt models, BERT embeddings and phonetic signals.

I also productionized our conversation dialog breakdown detection and human routing system (The Infinitus AI Machine blog post), and explored the use of latest prompt models for better performance.

Salesforce - Senior MTS

July 2018 - August 2022

Service Einstein

As a part of Einstein Agent, I worked on the no-downtime migration of ML apps (Case Classification, Case Wrap-up) to an advanced ML platform. I also led machine learning application health monitoring initiatives across Service Einstein teams that involved establishing SLI/O metrics, alerts, and dashboards for preemptive issue resolution.

I also developed a proof of concept for live chat summarization during case wrap-ups and architected the project’s pilot phase.

Field Service

As a part of the Advanced Preventive Maintenance team, I developed a robust system with an innovative DB schema for work order management of maintenance plans and assets with recurring maintenance schedules.

Internship: As a intern on the Field Service team, I developed an Entity Milestones Tracker Component for Lightning UI.

Oracle - Software Engineer

June 2014 - June 2016

As a part of the Core team of PeopleTools, Oracle PeopleSoft, I focused primarily on ensuring our software was WCAG 2.0 compliant, allowing complete screen-reader and keyboard navigation for various elements of PeopleTools application. I stabilized the platform in terms of accessibility for PeopleTools 8.55. I also provided sessions on accessibility to Tools and Apps teams to facilitate early inclusion of accessibility into product by developers. I was involved in development and maintenance of various core PeopleTools features

PayPal - Engineering Intern

January 2014 - June 2014

I worked in the uber-cool FPTI team (First Party Tracking Infrastucture), PayPal’s analytics platform.

Scalable Query Framework

This was my final year undergraduate research project - build a query framework to provide near-real time responses to queries over terabytes of PayPal user clickstream data. It would in turn help the Product Management make better and faster decisions for PayPal. We used Hadoop, HBase, and Elasticsearch to build the solution.

Real-time Analytics Dashboard

I worked on building the first visualization system using D3.JS, NodeJS and AngularJS to interface with Druid data-store of PayPal Analytics. The enhancement of this initial solution was presented in Strata+Hadoop world 2014 conference by the team at PayPal.

Research

Sequential Recommender Systems

September 2017 - June 2018

Master’s thesis under Prof. Julian McAuley on sequential recommender systems. I worked with metric embeddings to model user-item and item-item interactions in one metric space and incorporated various techniques to model temporal and geographical behavioural rhythms in human mobility.

Personalized Next Song Recommendation

October 2017 - December 2017

Modelling of user listening behaviour on Spotify and Last.fm by modelling song sequences in 30Music and NowPlaying datasets. We empirically prove the effectiveness of metric embeddings over matrix factorization and factorized markov chain models and suitably extend the embedding model.

Modeling the Evolution of User Expertise

March 2017 - June 2017

Using a latent factor model to model the evolution of user expertise by fitting a recommender system for each level of expertise. Project reproduces the results obtained in the main paper.

Teaching

Discrete Mathematics (undergraduate) Spring 2018

One of 6 teaching assistants for the undergraduate class on discrete mathematics under Prof. Miles Jones
Course: CSE 21

Algorithms (undergraduate) Winter 2018

One of 5 teaching assistants for a undergraduate algorithms class under Prof. Andrew Kahng
Course: CSE 101

Algorithms (graduate) Fall 2017

One of 5 teaching assistants for a graduate algorithms class of 187 students under Prof. Ramamohan Paturi
Course: CSE 202

Database Systems (undergraduate) Spring 2017

One of 4 teaching assistants for an undergraduate database systemsclass of 134 students under Prof. Yannis Papakonstantinou
Course: CSE 135

Academic Project Highlights

Las Vegas Restaurant Recommendation
Built a restaurant recommendation system using Bayesian Personalized Ranking (BPR-MF) for Las Vegas establishments, leveraging data from the Yelp Round 9 Dataset challenge.

Amazon Review Helpfulness and Rating Prediction
Developed prediction models as part of CSE 258 - Web Mining and Recommender Systems. Built regression models to predict helpfulness votes for product reviews and implemented matrix factorization techniques for product rating prediction.

Detection of Duplicate Question Pairs on Quora
Implemented Quora’s LSTM with concatenation architecture to detect semantic similarity between question pairs. Experimented with various deep learning architectures and word embedding techniques for optimal performance.

Bandit Convex Optimization
Authored a review paper for CSE 291 - Convex Optimization analyzing: various algorithms in Online Convex Optimization, bandit convex optimization techniques and algorithms promising tighter bounds for regret functions in bandit settings