Nov 2024 – Mar 2025 · 4 mos · San Francisco, California, United States · On-site

• Designed and implemented evaluation frameworks to test and validate the full AI pipeline.
• Improved entity search functionality to enhance search capabilities within Sweep’s AI coding assistant.
• Optimized end-to-end pipeline and search latency, reduced initial response times from over 40 seconds to under 1 second, improved
• search latency from over 30 minutes on large codebases to under a minute and improved the vector search implementation.
• Built key UI components for the Sweep JetBrains plugin, improving user interaction and workflow.
• Introduced file and entity mention features with intelligent prompt suggestions in the plugin.
• Implemented file syncing and developed the first version of in-editor search within the plugin

Jun 2023 – Oct 2024 · 1 yr 4 mos

• Developed and deployed initial machine learning models for image segmentation, classification, and movement detection, power-
• ing the functionality of the company’s nail painting robot.
• Built tools for efficient segmentation and keypoint annotation on 3D/2D images, accommodating multi-labeler collaboration.
• Built tools for interfacing with, calibrating, and controlling peripheral devices, and user interfaces for interacting with the controls.
• Contributed to multiple projects, including frontend development, calibration optimization, and implementing path planning and
• execution algorithms, significantly improving system usability and operational efficiency

Jan 2023 – May 2023 · 4 mos

• Teaching Assistant for the course Principles of Software Development

Jan 2023 – May 2023 · 4 mos

• Developed multimodal driver state estimation models integrating gaze tracking, hand gesture recognition, and biometric signals to assess driver safety for AI-driver control handover in autonomous vehicles.

Aug 2022 – Dec 2022 · 4 mos

• Teaching Assistant for the graduate level course Analysis of Algorithms

Jan 2022 – Jul 2022 · 6 mos

• Built computer vision models for object detection in a quadrupedal robot motion planning project

Jan 2022 – May 2022 · 4 mos

• Grader for the graduate level course Analysis of Algorithms

Sep 2021 – Dec 2021 · 3 mos

• Extract features for discovery and translation in Autism Spectrum Disorder by analyzing video data with Computer Vision models.

Jan 2023 – May 2023 · 4 mos

• Developed and optimized modules for neural network quantization, achieving minimal accuracy loss, tailored for the company’s proprietary Xtensa microprocessor.
• For instance, Mobilenetv3's accuracy with my quantization scheme was 70% compared to 66.66% with pytorch quantization on the xtensa microprocessor

Sep 2022 – Dec 2022 · 3 mos

May 2022 – Aug 2022 · 3 mos

• Built crawlers to extract several biomedical datasets.
• Extracted named entities in unstructured biomedical text and linked these to entities in structured medical databases
• Best model has a top-5 accuracy of 95.93% and top-1 accuracy of 59.62%
• Wrote several ensemble methods to improve top-1 accuracy to 61.19%
• Designed embeddings for both entities to allow zero-shot and few-shot learning

Dec 2020 – Jun 2021 · 6 mos

• Tech Lead in the search engineering department. Took on the role of tech lead in just a year of joining the firm.
• L2 lead in the search engineering department. Made sure all the code being pushed to the repository was clean, optimized and correct.
• Managed and mentored 1 Associate, 1 Senior Analyst and 2 Analysts
• Built a highly available, forever running, near-zero median latency, stable distributed search engine (Neon) on text, voice, and structured data, from scratch, supporting natural language queries over a few hundred terabytes of data.
• Tech stacks include Java, Python, Hadoop, YARN, MapReduce, Flink, Netty
• Built the core architecture and frameworks of the search engine
• Designed micro-services architecture with load balancing, high availability, and efficient container registration, discovery, and sharing
• Wrote NLP models for a Query Auto-complete service for the search engine on varied datasets. Achieved a 95% hit rate and under 25 ms latency on the model and increased the traffic on external-user facing datasets by 20%
• Built an extremely fast, low memory overhead spell correction algorithm for the autocomplete module
• Launched a search-summarization and result-clubbing module with efficient NLP models to help users find relevant content faster
• Improved latency by 92% with smart serialization schemes, I/O designs and caching mechanisms.
• Reduced memory usage by over 40% with better serialization and compression algorithms.
• Achieved ~100% up-time for external user facing data-sets
• Managed the release process for the search engine
• Achieved a 99.98% reduction in deployment time
• Made the search engine secure end-to-end and handled the tech risk for the search engine
• Built the initial parts of real-time data ingestion and retrieval pipeline
• Built several alerting and monitoring daemons for the search engine
• Built the query routing module
• Highly improved disk usage, cluster usage and swap memory usage

Dec 2018 – Dec 2020 · 2 yrs

• Tech lead in the search engineering department. As a part of my role, I build the frameworks on which the search engine runs, manage stability, networking, discovery, latency, security, and memory footprint of the search engine, build the query autocomplete and manage/mentor a team of 2. I also manage the deployment, release, and the core architecture behind the search engine

Jun 2017 – Dec 2018 · 1 yr 6 mos

• Tech lead in the search team
• Built the search engine for e-communications data
• Built indexing, UI, retrieval, query parsing, and ranking for transactions data
• Supported search on all order/execution entities along with real number entities with a below 2-second median latency
• Built the logging and metrics framework for the search engine

May 2017 – Jul 2017 · 2 mos

• Built a search engine(Neon) from scratch
• Launched first version of Neon which went on to become a separate team of its own with 40+ engineers, set to cater to all search requirements at Goldman Sachs.
• Built the initial parts of the architecture, indexing, query parsing, I/O, UI, ranking and retrieval

May 2015 – Jul 2015 · 2 mos · Bangalore

• Built the complete iOS application for the firm’s Social media assistant, DrumUp, which crawls the web and displays articles according to the user’s preferences.