Vinayak Sharma

[RA] Dec 2024 - Present Under Professor Prof. Aviral Shrivastava:

• Researching how to improve current NISQ Quantum Computers using SOTA Machine Learning

[TA] Jan 2025 - May 2025 CSE 598:Quantum Computing under Prof. Aviral Shrivastava:

• Created autograders for coding assignments in Pennylane
• Created final exams and quizzes

[TA] Aug 2024 - Dec 2024 486: Capstone Project under Prof. Helen Chavez:

• Helping final year undergraduate students with any issues they face in their ‘Capstone Project’

[TA] Jan 2024 - May 2024 CPI 111:Game Development I under Prof. Yoshihiro Kobayashi:

• Providing in-class support during game development.
• Working with a team of 2 graders and 5 UG-TAs.

[TA] Aug 2023 - Dec 2023 CSE 598:Quantum Computing under Prof. Zilin Jiang:

• Graded and provided feedback on assignments related to randomized classical algorithms, basics of qubits and quantum computing systems, quantum cryptography and QML.

[TA] Aug 2022 - Dec 2022 486: Capstone Project under Prof. Helen Chavez:

• Helping final year undergraduate students with any issues they face in their ‘Capstone Project’
• Teaching the fundamentals of ‘Agile Development’
• Giving guidance on the job search, interview and salary negotiation process

• Accepted to a three-month program that matches students with Google mentors and peers to support their pursuit of computer science research pathways. (research.google/outreach/csrmp)

At Myelin Foundry I worked on :
Aug 2020 - Jun 2022

• Overhauled the inference pipeline for a multi-model orchestration system comprising face detection, recognition & driver management models, resulting in a 70% increase in the number of concurrently running models.
• Optimized computer vision system utilizing Tensorflow with TensorRT acceleration on Nvidia Jetson platform, resulting in a 2x increase in inference speed and 50% increase in defect detection rate in real-time steel manufacturing.
• Reduced footprint of a low-light image enhancement model by 50% by adapting the ZeroDCE++ architecture to run faster on a DSP enabling deployment on embedded devices.
• Created a new architecture and training pipeline for a joint denoise & super-resolution (SR) model reducing compression artifacts by 60% for Super Resolution outputs on encoded video streams.
• Awarded the Myelin Impact Award as the most impactful employee for the company in the year 2021.

Jun 2019 - Jul 2019

• Created an end-to-end speech-to-text system for Pashto - a challenge due to the extreme data scarcity associated with the language
• Tested & evaluated the viability of Tensorflow.js for deployment of the organization’s product stack
• Helped set up the company servers for model training

As part of a team of student researchers under the Samsung PRISM program, I worked on:

• Online Handwriting Recognition in Bengali and Tamil.
• The dataset was collected by us on campus and comprised of 1300 handwriting samples.
• The data was in the form of XML Files with (x,y,t) information.
• The model we chose to use was an attention-based LSTM to learn more information.
• This model was combined with a pre-trained language model - BERT to improve the accuracy of our predictions.

As an intern for Tata Elxsi, I was tasked with :

• Creating an API to perform object detection in identifying humans from video footage and estimating their distance.
• One of the many proposed applications for the API was to enforce Social Distancing protocols to assist clients better ensure safe working conditions.
• The API was packaged into a Docker container to maintain stability and ease of sharing.