Yashal Kanungo

AI Researcher

Seattle, Washington, United States9 yrs 8 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Led generative AI research at Amazon Ads.
Developed multi-agent orchestration for ad generation.
Granted multiple patents in generative AI technologies.

Stackforce AI infers this person is a leader in Generative AI for Advertising.

Contact

Skills

Core Skills

Generative AiLarge Language Models (llm)Video GenerationDiffusion Models

Other Skills

Agentic systemsCross-functional Team LeadershipImage GenerationConditional Image GenerationText-to-Image GenerationStable DiffusionDiffusionArtificial Intelligence (AI)Machine LearningPythonDeep LearningCloud ComputingBayesian OptimizationAlgorithmsScala

About

I lead generative AI research and production applications at Amazon Ads AI Gen org. With 10+ years of AI/ML experience, including 6 years specializing in generative language models and image generation at scale, I bring rare end-to-end expertise in training and deploying both diffusion models for image generation/editing and LLMs for text generation. I've also built and led the launch of a multi-agent chat system with large-scale multimodal tool integration. I also have several years of experience managing the entire AWS infrastructure of the science team (GPUs, petabytes of data etc.). Previously, I led projects in ranking, classification, and retrieval of items at 100+ million scale, delivering systems that process content that gather billions of impressions. I also have some patents & publications in generative AI.

Experience

Amazon

7 roles

Lead Applied Scientist

Promoted

Nov 2024 – Present · 1 yr 4 mos

Leading generative AI research and productionization for Amazon Ads' AI Gen organization. Driving end-to-end innovation across text, image, and video generation for advertising, with systems creating content with Billion scale impressions.
Designed and built the multi-agent orchestration architecture for the Creative Agent — a conversational AI agent that generates studio-quality video, image, and audio ads from natural language prompts. Featured at AWS re:Invent 2025 and covered by Wall Street Journal, CNBC, TechCrunch etc.

Agentic systemsCross-functional Team LeadershipLarge Language Models (LLM)Video GenerationGenerative AI

Senior Applied Scientist

Promoted

Dec 2022 – Nov 2024 · 1 yr 11 mos

Led Amazon Ads' first generative AI (Diffusion models) image and video generation systems from research through production deployment across US, UK, EU, India, Canada, and Mexico — the first-ever commercial deployment of diffusion models for retail advertising at this scale.
Granted 3 US patents for video generation and image prompt technologies.

Generative AIDiffusion modelsImage Generation

Applied Scientist II

Apr 2021 – Dec 2022 · 1 yr 8 mos

DL NLU and NLG models for advertising.amazon.com
Basic TTS modeling
DL Hosting / Inference
Multi-modal image and video models

Applied Scientist I

Feb 2019 – Mar 2021 · 2 yrs 1 mo

Ranking model and sales prediction for deals.amazon.com
Ad Classification for advertising.amazon.com
Approx. NN model and keyword expansion model for discovery for advertising.amazon.com
Ad Sourcing model for an MVP for advertising.amazon.com

Research Scientist I

Promoted

Apr 2018 – Jan 2019 · 9 mos

• Ad Classification for advertising.amazon.com

Software Dev Engineer I (Machine Learning)

Aug 2017 – Apr 2018 · 8 mos

• Ad Ranking with contextual Click (CTR) and Conversion prediction (CVR) for Native Shopping Ads on a billion+ impression scale.

Software Dev Engineer I - Intern (Machine Learning)

Jan 2017 – Jul 2017 · 6 mos

• Advertisement supply classification and understanding model. Prediction of Product Centricity of web pages (NLU) on a billion+ impression scale.

Robert bosch engineering and business solutions limited

Machine Learning Intern

Aug 2016 – Jan 2017 · 5 mos

Prediction of Diabetic Retinopathy using CNNs in a low resource setting with images taken using new hand-held camera. Improved recall by 10% using transfer learning and label augmentation.
Worked with various ophthalmologists to validate the model
Improved model inference throughput by 10x and reduced cloud compute costs.
Worked on model inference for health camps for under-represented communities.
Experimented with alternate models and baselines and mobile prediction