Suvaditya Mukherjee

Suvaditya Mukherjee

ML @ USC Institute of Creative Technologies & USC School of Cinematic Arts | MS CS-AI @ USC Viterbi | Google Developer Expert (Machine Learning)

University of Southern California

USC Institute of Creative Technologies

USC School of Cinematic Arts

NMIMS University

Biography

He is currently working at USC Institute of Creative Technologies and the USC School of Cinematic Arts, while pursuing a Master of Science degree in Computer Science with a specialization in Artificial Intelligence from the USC Viterbi School of Engineering, Thomas Lord Department of Computer Science. He is also a Google Developer Expert for Machine Learning.

Interests
  • Machine Learning
  • Computer Vision
  • Deep Learning
  • Natural Language Processing
  • Generative Modelling
Education
  • Master of Science in Computer Science - Artificial Intelligence, 2026

    University of Southern California

  • B.Tech in Artificial Intelligence, 2024

    NMIMS University

Experience

 
 
 
 
 
USC Institute for Creative Technologies
ML Student Worker
September 2024 – Present Los Angeles, CA, USA

◦ Course Generation using Generative AI: Working under Prof. (Dr.) Benjamin Nye to help create novel techniques for course generation, tutoring content generation, and OpenTutor courses for learning AI using Generative AI for the AIRCOEE program in collaboration with the US Navy.

◦ Cogeneration Testbed: Helping maintain the technologies for co-generation of tutoring content using open and cloud-based LLMs.

 
 
 
 
 
USC School of Cinematic Arts
ML Assistant
September 2024 – Present Los Angeles, CA, USA

◦ Student Worker: Working under Prof. (Dr.) Mark Bolas to help create the introductory Python Programming course for Game Developers.

◦ ML Research: Helping to find new approaches to apply Generative AI using LLMs and Diffusion Models towards solving problems at a large-scale in the field of Creative Media like games and movies.

 
 
 
 
 
Harman International
ML Intern
December 2023 – May 2024 Remote, Bengaluru, India
◦ K-Shot Rotation-Invariant Object Detection Pipeline Development: Producing new Intellectual Property towards achieving a robust pipeline that can perform K-shot object detection without dependence on rotation alignment. Resulted in pipeline with 35% better results on client data ◦ Zero-shot Time-Series Forecasting with LLMs: Researching on how to perform zero-shot time-series forecasting using LLMs while building on previous developments in the space. ◦ Spot Instance Handler using Agentic LLMs: Created an agent-based LLM system using Gemini 1.5 Pro and LangChain that helped in reducing costs by 10% incurred from running non-critical workloads on spot-instances
 
 
 
 
 
Weights & Biases
Author
October 2023 – Present Remote, San Francisco (CA), United States of America

WandB Reports: Creating high-quality Weights & Biases Reports around the newest trends in Computer Vision and Machine Learning

Integrations: Developing new integrations for Weights & Biases with popular open-source libraries with logging capabilities

 
 
 
 
 
PyImageSearch
External Author
June 2023 – Present Remote, New York City (NY), United States of America

Blog Posts: Writing highly technical and researched blog posts around different challenges within Computer Vision and Machine Learning

Video Blogging: Developing informative videos around the written blog posts for generating educational content

 
 
 
 
 
Center for Visual Information Technology, IIIT-Hyderabad
Research Intern
June 2023 – October 2023 Hyderabad, Telangana, India

Research: Working towards a publication along Domain Adaptation problems in Autonomous Driving under Prof. C.V. Jawahar and Prof. Shankar Gangisetty

Code Implementations: Operating with internal tools to perform large-scale GPU training and experimentation on Image Segmentation problems

 
 
 
 
 
Ivy (unify.ai)
ML Research Engineer Intern
January 2023 – July 2023 Remote, London, United Kingdom

Demos and Examples: Work done towards adding newer demos, examples, and guides to the internal and external official documentation, most notably around converting torchvision models into TFLite

Front-end and Back-end APIs: To work on adding to existing infrastructure for front-end and back-end functions in TensorFlow and JAX

Internal AI Developer: To work on an AI Developer (Code-LLMs) that automates and builds upon the existing codebase and speeds up internal development, along with handling their training through Cloud resources like GCP and AWS

 
 
 
 
 
Mosaic Wellness
Software Engineer Intern
May 2022 – July 2022 Mumbai, Maharashtra, India

Image Annotation Tool: In-house end-to-end tool, developed with TypeScript (APIs from/into Amazon RDS) and integrated into a React frontend to be used by doctors for tagging and annotation of patient images.

ML CI/CD Pipeline: Designed and partially implemented a Continuous ML pipeline with use of AWS SageMaker, AWS RDS and AWS EC2 for deployment of a classifier.

Projects

*
TranscribeMate
A simple application developed in Flutter that enables transcribing information from images using Google MLKit For Flutter
TranscribeMate
Temporal Latent Bottleneck - TensorFlow
This is an unofficial implementation of the paper “Temporal Latent Bottleneck - Synthesis of Fast and Slow Processing Mechanisms in Sequence Learning” by Didolkar et. al., developed jointly by Aritra Roy Gosthipaty and Suvaditya Mukherjee
Temporal Latent Bottleneck - TensorFlow

Recent & Upcoming Talks

Contact

Reach out to me! I enjoy having conversations around Machine Learning, Computer Vision, Production Systems, and Google Cloud