Vashisht Madhavan
Break into AI | AI Researcher @ Amazon | Ex-Uber AI & YC Founder
Studied at University of California, Berkeley
Works at Amazon
Available tomorrow at 5:00 PM UTC
Questions? Start chatting with this coach before you get started.
Vashisht's Offerings
Custom hourly · $100/hr
Get help with AI Agents, AI Fundamentals, and .
Vashisht also coaches for Build with AI and Career Coaching. View all.
Vashisht’s Break Into AI Careers Qualifications
Experience level: Director
Welcome! I'm a Senior Applied Scientist at Amazon (AGI Foundations), where I work on post-training and reinforcement learning for code-generation models. Before Amazon, I built text-to-video generation models at Pika (the post-training behind features like Pikaffects), co-founded and was CTO of Humanlike (YC S23), led document-AI products at Snorkel AI used by Fortune 500 companies, and did deep reinforcement learning research at Uber AI Labs — with published work at NeurIPS, ICLR, and CVPR and a master's from UC Berkeley. I've also sat on the hiring side at both startups and big tech, so I know exactly what AI teams screen for. Whether you're a student, a career-changer, or an engineer pivoting into AI, I'll help you pick the right lane (research vs. applied vs. engineering vs. product), build portfolio projects with the tools employers actually look for (PyTorch, Hugging Face, fine-tuning, RAG, LLM pipelines), position your resume and projects so they land, and prepare for technical and behavioral interviews. I've navigated a non-traditional path through research, startups, and big tech myself — let's connect and chart your path into AI!
Vashisht can help with:
AI Agents
AI Fundamentals
Cover Letters
Interview Prep
Job Search Strategies
Resume
Salary Negotiation
Skill Building
Vashisht also coaches for Build with AI and Career Coaching. View all.
Work Experience
Senior Applied Scientist
Amazon
January 2025 - Present
At Amazon, I'm an AI researcher working on LLMs for coding. I've been responsible for helping post-train our latest model release, Nova 2 Pro. My focus is on improving how these models reason and write code through reinforcement learning.
AI Researcher
Pika
January 2024 - January 2025
At Pika, I was a researcher working on the company's generative AI models for video. I led data curation for large-scale pretraining and owned much of the post-training stack. That work powered a range of user-facing fine-tunes like Pikaffects, and I also fine-tuned multimodal LLMs for video captioning, categorization, and retrieval.
Co-Founder & CTO
Humanlike
May 2023 - December 2023
At Humanlike (YC S23), which I co-founded and ran as CTO, I built low-latency, realistic AI voice agents that automate business phone calls using LLMs and generative speech models. I led engineering and product end to end, from the first prototype to landing enterprise customers. We went through the YCominbator S23 batch.
Staff Machine Learning Engineer
Snorkel AI
July 2021 - May 2023
At Snorkel AI, I was a staff ML engineer and tech lead for the structured-documents product, which let enterprises extract, label, and classify data from PDFs and other documents. I brought foundation-model workflows into the platform — embedding search, prompt-based labeling, and LLM fine-tuning with document-specific models like LayoutLM — alongside integrations with Snowflake, Azure, and Databricks.
AI Researcher
Uber
May 2017 - November 2019
At Uber, I was a research scientist in Uber AI Labs, the company's artificial-intelligence research group. I worked on reinforcement learning and computer vision, specifically for self-driving cars and to optimize Uber Freight surge pricing. Was also involved in building an AI customer service agent that saved the company tens of millions of dollars annually.
Education
University of California, Berkeley
Master of Science (M.S.), Electrical Engineering and Computer Science
2016 - 2017
AI and Machine Learning. Focus on computer vision and reinforcement learning
University of California, Berkeley
Bachelor of Science (B.S.), Electrical Engineering and Computer Science
2012 - 2016
1 Review
Overall Rating
5.0
Knowledge
5.0
Value
5.0
Responsiveness
5.0
Supportiveness
5.0