- Researching and developing infrastructure for DSO.ai, a Synopsys tool that searches foroptimization targets in very large solution spaces of chip design, utilizing reinforcementlearning to enhance power, performance, and area- Enhanced data harvesting from unstructured log files and generated time series metricgraphs to ease in identifying steps that improved/degraded the performance of a run.- Reduced time-to-result in inference by pruning hyperparameters selected via fANOVA.- Designed execution engine infrastructure to replay crashed sessions and ensure lost metricinformation is recovered.- Overhauled the logging infrastructure thus improving customer experience by generatingconcise, readable logs.- Mentored an intern over two summers and managed their project end-to-end. Onboardedthem to our codebase, defined key milestones, and cleared roadblocks along the waythrough training.Skills: Machine Learning, System Design, MentorshipTools and languages used: Python, DSO.ai, Fusion Compiler, Perforce

- Developed, tested, and maintained code for the Calendar and Email service in Alexa. - Implemented a unified OAuth-based account linking flow for Calendar and Email in Europe and Asia through the Alexa mobile app to enable Calendar notifications.- Implemented Calendar account linking through Alexa Routines resulting in an increase in the customer base by about 500 customers per day. - Instrumented BI metrics in the Alexa app to quantify customer engagement with the service in the Alexa mobile app.

- Implemented multi-object tracking using the DeepSORT tracking algorithm- Classified vehicles by color using the video feed- Automated testing capabilities for new models on a Jetson NanoTools and languages used: TensorFlow, Python, AWS, Git

- Designed a physics-informed accident prediction AI with dashcam videos as input- Extracted physical priors - optical flow, distances, and velocities - from monocular videos to boost the performance of traditional neural networks by blending physics with deep learning- Applied adversarial attacks on the network to make it robust and explainableLab: Visual Machines GroupSupervisor: Prof. Achuta KadambiTools and languages used: Python, PyTorch, Keras, TensorFlow

- Developed an online EEG-based Brain-Computer Interface (BCI) using code-modulated visual evoked potentials (cVEP) and beamforming- Designed and conducted experiments with human participants to analyze the effectiveness of the BCI- Achieved 97.40% average accuracy in classification with 32 targets, and maximal Information Transfer Rate of 85.20 bits/minLab: KU Leuven - Laboratorium voor Neuro- & PsychofysiologieSupervisor: Prof. Marc M. Van HulleTools and languages used: Psychtoolbox, MATLAB

- Tested and verified a RADAR-based obstacle detection system for automobiles- Designed a visualization tool for real-time obstacle positions and velocitiesTeam: Advanced Driver Assistance Systems (ADAS)Tools and languages used: C, Gnuplot

- Investigated methods Radio Frequency Interference (RFI) detection in signals received by radio telescopes- Implemented the Spectral Kurtosis algorithm to identify the frequency bins containing RFI- Obtained 99.16% precision in detection Lab: Statistical Signal Processing Laboratory, Department of Electrical Communication EngineeringSupervisor: Prof. K.V.S. Hari Language: MATLAB