Vivek Vijaykumar Email & Phone Number

San Francisco, CA, US

Lakehouse Monitoring, AI Data Quality Team

Atlanta, Georgia, US

Graduate TA for CS 6476 (Introduction to Computer Vision) under Prof. Judy Hoffman

Atlanta, Georgia, US

• Researcher in Prof. Judy Hoffman’s computer vision and machine learning lab.
• Developing techniques to improve synthetic to real domain generalization and adaptation for semantic segmentation, object detection, and classification.
• [ICCV 2023] Co-authored Publication: "PASTA: Proportional Amplitude Spectrum Training Augmentation for Syn-to-Real Domain Generalization" (arXiv:2212.00979)
• [TMLR 2024] Co-authored Publication: "We're Not Using Videos Effectively: An Updated Domain Adaptive Video Segmentation Baseline" (arXiv:2402.00868)

Atlanta, Georgia, US

TA for CS 4476 (Introduction to Computer Vision) for Prof. Judy Hoffman and Prof. James Hays

San Francisco, CA, US

• ML Data Team
• Developing Lakehouse Monitoring Product

Menlo Park, CA, US

• Messenger Kids Core Experiences Team
• Engineer on ML Growth team for Messenger Kids; Accomplished all project and stretch deliverables ahead of schedule.
• Developed internal tool to help debug and provide end-to-end visualization of current Multi-Task Multi-Label Ranking Models used in production that impact millions of users; Tool helps shape roadmap of team when.
• Investigated and presented findings on simulator methods to obtain meaningful offline metrics on ML models.
• Explored and developed new ML model architectures to improve candidate targeting and ranking

Seattle, Washington, US

• Content Creation and Optimization Team
• Project goal was to build generative AI/ML models that could streamline the studio shots workflow, help reduce time to market for new clothing lines, and enable virtual try-on methodologies.
• Sole individual responsible for end-to-end project delivery: ideation (built from ground-up using techniques outlined in recent papers), architecture design, data collection, model training, deployment (RESTful API).
• Demo-ed to 50+ people in project showcase.

Berkeley, CA, US

• Freshman researcher in Prof. Kurt Keutzer’s lab; BAIR is one of the top AI research groups in the country.F͟a͟s͟t͟ ͟M͟o͟c͟o͟:͟
• Research and implement techniques to speed up training for Unsupervised/Self-Supervised Representation Learning.
• Implemented techniques including 1cycle and multi-cycle training, progressive image sizing, mixed-precision training on PyTorch codebase utilizing state-of-the art methods from recent publications including Momentum.
• Contributed to and performed experiments for work on HPT for Self Supervised Learning
• Investigated robustness and fast convergence qualities of HPT relating to augmentation sets, amount of pretraining data, and selected basetrain models used for self-supervised pretraining.
• Co-authored Publication accepted at WACV conference: "Self-Supervised Pretraining Improves Self-Supervised Pretraining" (arXiv:2103.12718);

Dearborn, Michigan, US

• Project goal was to use computer vision to enable scooters to self-park themselves at charging stations after being left in streets by customers, reducing costs and improving accessibility of fully-charged scooters.
• Designed and implemented a 3D visual target tracking system to identify charging stations, and a finite state machine to dictate movement path for the scooter to dock itself. Submitted 3 patent invention disclosures.
• Won 1st prize at Ford Hackathon by building an ML model that used images and data collected from Ford vehicles to predict road-quality in real-time; Proposed application to be used by cities to have a real-time map of.

Master'S Degree, Computer Science

Bachelor Of Science - Bs, Computer Science

High School