Sahil Goyal

Product Engineer

Bathinda, Punjab, India6 yrs experience

AI EnabledHighly Stable

Key Highlights

First-author publications at top AI conferences
Best Demo Paper award at ACM MM Asia '22
Experience with leading ML projects and hackathons

Stackforce AI infers this person is a Machine Learning and AI specialist with a focus on generative models and optimization.

Contact

Skills

Core Skills

Machine LearningDeep LearningResearchComputer VisionOptimization

Other Skills

Artificial Intelligence (AI)CNNCompetitive ProgrammingData ScienceEfficient and Elastic Image/Video GenerationFlaskGeminiGenetic AlgorithmsGitGitHubHackathonsHeuristic algorithm optimizationHuman Computer InteractionMathematicsModel merging and compression

About

I am a Research Associate at Google DeepMind (Frontier AI Unit) in MLO (Machine Learning Optimization) team. I completed my Integrated Masters in Mathematics and Computing at IIT Roorkee with department Gold Medal. Currently, I am working on Efficient and Elastic generative models. During my undergraduation, I did internship at Adobe Research, where I worked on evaluation/refinement of graphic designs with Dr. Joseph KJ and Dr. Balaji Vasan Srinivasan. I also interned at MIDAS Lab, where I worked on talking face generation with Dr. Rajiv Ratn Shah. My research interests include Generative Models, Efficient ML, and broad Deep Learning topics in general. I have first-author publications at ICML, WACV, ACM MM Asia, ICLR (W), ICCV (W), ACM MM (W) conferences. I was awarded Best Demo Paper award at ACM MM Asia '22. I have also served as reviewer in conferences like CVPR, ICML, ICLR, WACV, AAMAS multiple times.

Experience

Google deepmind

Research Associate

Jun 2024 – Present · 1 yr 9 mos · Bengaluru, Karnataka, India · On-site

Efficient and Elastic Image/Video Generation
Model merging and compression
Working on VEO and Gemini

Efficient and Elastic Image/Video GenerationModel merging and compressionVEOGeminiMachine LearningDeep Learning

Adobe

Research Intern

May 2023 – Jul 2023 · 2 mos · Bengaluru, Karnataka, India · On-site

A novel self-supervised approach to evaluate Graphic Designs quantitatively (Layout, Color, and Typography).
Used Siamese networks along with Design Heuristics to score Graphic Designs.
Modified advanced Genetic Algorithms (NSGA-II) for Layout optimization.
Deployed the entire system as a Flask Web Application.
Paper accepted at WACV '25

Self-supervised learningSiamese networksGenetic AlgorithmsFlaskComputer VisionMachine Learning

Midas: multimodal digital media analysis lab

Research Intern

Sep 2021 – Jan 2023 · 1 yr 4 mos

A novel CNN-based talking face generation network involving lipsync and explicit emotions.
New skip-connections-based concatenation technique introduced.
SOTA results on CREMA-D dataset ( 83.2% emotion classification accuracy and FID metric: 5.29 ).
Deployed the entire system as an interactive web interface.
Papers accepted at ICCV '23 (W) , ACM MM '23 (W) , and ACM MM Asia '22.
Worked under the supervision of Dr. Rajiv Ratn Shah.

CNNTalking face generationSkip-connectionsWeb interface deploymentDeep LearningComputer Vision