# Rhythm Garg - Applied Compute Source: https://www.youtube.com/watch?v=1ZHPrqXpbBk Recap page: https://rapidrecap.app/video/1ZHPrqXpbBk Generated: 2026-02-04T22:32:01.949+00:00 --- ## Quick Overview Applied Compute, co-founded by Rhythm Garg, focuses on helping enterprises build and own internal, unique AI workforces by leveraging cutting-edge research and rapid deployment, exemplified by a successful partnership with DoorDash that reduced erroneous menus by approximately 30%. **Key Points:** - Applied Compute's mission is to help enterprises build and own internal, unique AI workforces powered by proprietary data and tailored to specific workflows, enabling continuous improvement over time. - The company's approach is Product-Led, Research-Enabled (using frontier tools), and Forward-Deployed (to ship real outcomes). - The co-founders, including Rhythm Garg and Yash, came from backgrounds involving early research at places like OpenAI and working on the RL stack at OpenAI. - A key success story is a custom merchant onboarding system deployed at DoorDash, which resulted in a ~30% relative reduction in erroneous menus across full traffic. - The team has experience from top-tier AI/tech companies including OpenAI, Scale, Meta, NVIDIA, Databricks, and Google, highlighting their deep technical expertise. - Applied Compute is actively hiring for Product, Research Systems, and Applied Research engineers to build infrastructure, power frontier research, and deploy models. - The company culture emphasizes enjoying the work, being close friends, and working hard to lead the transformation toward enterprise-specific intelligence. ![Screenshot at 0:05: The initial slide contrasts the observed trend that 'agents are getting smarter' with the reality that 'agents fail in the enterprise,' setting up the problem Applied Compute aims to solve.](https://ss.rapidrecap.app/screens/1ZHPrqXpbBk/00-00-05.jpg) **Context:** Rhythm Garg, Co-Founder and CTO of Applied Compute, presents the company's core mission and approach during a talk, likely at a startup or AI conference. The presentation outlines the challenges of deploying advanced AI agents in the enterprise, the company's background stemming from deep RL research at OpenAI, its key partnership success with DoorDash, and its hiring strategy, all supported by slides displaying company affiliations and team photos. ## Detailed Analysis Rhythm Garg introduces Applied Compute, a company focused on enabling enterprises to build and own internal, unique AI workforces that leverage proprietary data and are continuously improving. He notes that while AI agents are rapidly advancing, out-of-the-box agents frequently fail in real-world enterprise automation tasks. Garg highlights his background, having worked with co-founder Yash on the RL stack at OpenAI, where they saw rapid advancements weekly, which fueled their belief that transformative change was possible. Their core problem focus is realizing billions of dollars in value by solving enterprise scalability issues. The company's approach is threefold: Product-Led (to deploy faster), Research-Enabled (to use frontier tools), and Forward-Deployed (to ship real outcomes). They build state-of-the-art RL stacks that allow customers to rapidly deploy models and accelerate deployments. As a concrete example, Garg details a collaboration with DoorDash where their custom merchant onboarding system was deployed to production handling full traffic, leading to a significant ~30% relative reduction in erroneous menus. This work involves converting unstructured menu data into structured DoorDash objects, utilizing their internal tools and domain expertise. The team's background is impressive, having grown up in leading AI organizations like OpenAI, Scale, Meta, NVIDIA, Databricks, and Google, indicating deep technical pedigree. Garg emphasizes that the team is driven by shipping specific intelligence to the enterprise, not just general AI. Finally, he notes that they are a small, growing team actively hiring across Product, Research Systems, and Applied Research roles, fostering a culture where they have fun while working hard. ### Initial Problem Statement - Agents are getting smarter, but agents fail in the enterprise - Solving this unlocks billions of dollars in value - Focus is on scaling enterprise automation. ### Applied Compute Approach - Product-Led (deploy faster) - Research-Enabled (use frontier tools) - Forward-Deployed (ship real outcomes) - Building state-of-the-art RL stacks for continuous improvement. ### Case Study - DoorDash - Custom merchant onboarding system deployed to production over full traffic - Achieved ~30% relative reduction in erroneous menus - System automates menu conversion using internal data and expertise. ### Team Background - Team members originated from OpenAI, Scale, Meta, NVIDIA, Databricks, and Google - Co-founders previously worked on RL stack at OpenAI - Driven by shipping specific intelligence to the enterprise. ### Company Culture & Hiring - Small, growing team of under 20 people looking to expand - Culture values having fun and working hard - Hiring for Product, Research Systems, and Applied Research roles. ![Screenshot at 0:00: Rhythm Garg presenting on stage with dual screens showing the Applied Compute title slide.](https://ss.rapidrecap.app/screens/1ZHPrqXpbBk/00-00-00.jpg) ![Screenshot at 0:05: Slide showing the core dichotomy: "agents are getting smarter" leading to "agents fail in the enterprise."](https://ss.rapidrecap.app/screens/1ZHPrqXpbBk/00-00-05.jpg) ![Screenshot at 0:34: Slide titled "About us" displaying logos of former organizations including Stanford, O1, and the Applied Compute logo.](https://ss.rapidrecap.app/screens/1ZHPrqXpbBk/00-00-34.jpg) ![Screenshot at 1:04: Slide stating the mission: "We help enterprises build and own an internal, unique AI workforce."](https://ss.rapidrecap.app/screens/1ZHPrqXpbBk/00-01-04.jpg) ![Screenshot at 1:22: Slide detailing their approach: "Product-Led," "Research-Enabled," and "Forward-Deployed."](https://ss.rapidrecap.app/screens/1ZHPrqXpbBk/00-01-22.jpg)