Feb 2025 – Apr 2025 · 2 mos

Jun 2023 – Feb 2025 · 1 yr 8 mos

• Leading a team of 9 engineers across two time-zones spanning three core services within the company.
• The team is responsible for development and maintenance of natural language search functionality in our analytics software.
• Led the ideation and implementation of feedback (human in the loop) incorporation mechanism to improve accuracy of Natural Language Search (95% reported by customers) & ensure customers have full control over training their queries using the system. The feedback is multi-level and handles database queries, chart configurations, formula and join specifications of users enabling seamless personalization for all users in a cluster
• Implemented adhoc formula & query on query generation using natural language where inference is made based on SQL from LLM to decide if we need to generate a formula to create the AI generated answer.
• Led work on developing multi-level semantic cache to reduce costs & latency associated with API calls to LLM services.
• Working with metadata teams to move single tenant services to multi-tenant as part of the firm's NorthStar architecture.

Feb 2022 – Jun 2023 · 1 yr 4 mos

• Responsible for maintaining & optimizing metadata infra.
• Reduced memory consumption of metadata service by 4x, bringing down java Xmx from 64GB to 16GB in production environments, reducing cloud costs for the firm.
• Developed NL to DSL using a combination of LLMs & heuristics with an accuracy of 73% on worksheets with 1000s of columns.
• Conceptualized & developed online human feedback sourcing & incorporation algorithm to improve NL to DSL translations. (Patent Pending)

Feb 2021 – Jan 2022 · 11 mos

Feb 2020 – Jan 2021 · 11 mos

Feb 2019 – Jan 2020 · 11 mos

• Part of the Search team at ThoughtSpot.
• Responsible for developing search algorithms for performing queries on large datasets using natural language for data analysis.
• One of the core engineers who built a multi-tenant scalable search system using Apache Kafka, Elasticsearch & Kubernetes, from concept to final implementation. Link: https://www.thoughtspot.com/product/search

Oct 2018 – Dec 2018 · 2 mos · Greater New York City Area

Jun 2018 – Aug 2018 · 2 mos · Greater New York City Area

• Worked on migrating legacy applications, setting up build environments, automated build process, developed applications to generate audit check reports, notifying credit event to stakeholders & modified replenishment logic based on new trades
• Synchronized servers within avg. 7 µs accuracy (10% network load) using SVM hyperplane intercept as offset b/w servers
• Developed an email classification plugin in C# to improve efficiency by reducing spam but allowing anomaly notification

Jan 2018 – May 2018 · 4 mos

• Worked on solving word algebraic problems using Structured Prediction Energy Networks

Mar 2017 – Aug 2017 · 5 mos · New Delhi Area, India

• Designed & built end to end system for contextual content suggestion including data collection & analysis, model training & parameter tuning, & running neural network algorithms on smartphones using C++ (Core) & Java (Android)
• Achieved 139% increase in daily average content shares & 105% increase in average user sessions created in Bobble keyboard
• Developed transliteration & word prediction algorithms using bidirectional LSTMs with attention & trigram models for Indic languages improving transliteration user accuracy from 69.89% to 76.96% & test accuracy from 59% to 74%
• Reduced latency of image processing & neural network models on smartphones by over 7 times using C++/OpenGL

Jul 2015 – Dec 2016 · 1 yr 5 mos · Noida Area, India

• Responsible for feature prototyping, code analysis, testing & optimization of settings & gallery modules in smartphones
• Implemented a prototype for automatic categorization of images into documents, objects, landscapes, selfies & group photos in gallery application using an ensemble of pretrained image classification models

May 2014 – Jul 2014 · 2 mos · Bengaluru Area, India

• Development of User Context Detection System using Smartphone Sensors