Tanmay Kulkarni Email & Phone Number

Mountain View, California, United States

Pixel Watch AI

Pittsburgh, Pennsylvania, United States

• Developed infrastructure for Autonomous Vehicle (AV) behavior verification through comprehensive offline testing, reducing verification time from several hours to mere minutes.• Engineered an automated labeling system to rapidly identify and tag bad behaviors in AVs, significantly accelerating the process from a lengthy manual task to a few seconds.

Pittsburgh, Pennsylvania, United States

• Experimented on different transformer based backbones for Semantic Segmentation• Researched and analyzed different approaches for unsupervised domain adaptation• Applied prompt engineering to language and image-text models

Pittsburgh, Pennsylvania, United States

• Implemented model based learning algorithms for reinforcement learning• Combined model based approaches along with action primitives to improve sample complexity of the approach• Experimented with the use of Wrist-Camera for additional supervision, Hindsight Experience Replay and different vision encoders

Bengaluru, Karnataka, India

• Deployed event based concurrency model for ensuring consistency across Teams meetings• Developed controller for Directory Service API which handles batching and parallel requests with an 99.9% availability• Configured and deployed monitors to test the availability of deployed services• Developed a HTTP Retry Client to improve the stability of service to service communication

Bangalore

• Created a framework with scikit-learn, pandas, numpy and python to create adversarial datasets for common estimation problems• Designed and implemented refuters which challenged the basic assumptions while constructing a statistical model corresponding to a dataset.• Executed experiments for a combination of datasets, models and refuters on High Performance Computing Clusters with the help of Docker and PowerShell.

India

• Created an end-to-end machine learning pipeline for deploying Object Detection Models on TensorFlow Lite• Collected data from multiple sources such as Pascal VOC, COCO, and the Open Images Dataset• Performed data cleaning and converted the data to the TFRecord Format.• Trained multiple deep learning models such as MobileNet SSD, YOLO and Faster RCNN• Created a basic interface for deploying the models with the help of an Android Application

Bangalore

• Ported Document Library within Microsoft Teams to ReactJs from AngularJs• Integrated telemetry to monitor performance of service• Implemented to end-to-end testing of scenarios

Hyderabad

• Teaching Assistant for Discrete Structures under Professor Bhanu Murthy• Involved in creating problem sets, organizing labs and invigilating tests

Bengaluru Area, India

• Scrape the data from multiple search engines with the help of Python and Selenium• Aggregate the data for the SmartSearch application to get more relevant results

Bengaluru Area, India

• Create a new interface between the LEGO Mindstorms Kit and a more powerful processing unit such as a Desktop or Mobile Phone• Explored the trade-off between a higher latency but more powerful chip present remotely to having a lower latency, less powerful chip present on the Robot• The interface was implemented with the help of MATLAB and Java

Master Of Science - Ms, Computer Science

Master'S Degree, Economics

Bachelor'S Degree, Computer Science