Product Manager, AI Research

About this role

<p>Descript’s vision is to put video in every communicator’s toolkit. Back in the day you needed like six monitors and a bachelor’s degree to edit video. Descript lets you do it by editing docs & slides, and increasingly by just asking AI. In the future, maybe you won’t even need to ask! But building a new way to record or generate (or both!) videos that look & sound good comes with a series of unique design, technology, and business challenges. In other words, we need really good product managers.</p> <p>We’re looking for a Product Manager to help build the future of video editing with AI. You’ll work alongside a small, flat, highly collaborative team of experienced PMs, AI researchers, engineers, designers, and marketers. This is an opportunity to get hands-on experience with cutting-edge AI technology in a product users love and grow fast in your PM craft.</p> <p>We're looking for a Product Manager to lead the AI Research and Enablement roadmap at Descript. This role sits at the intersection of cutting-edge AI research, production ML infrastructure, and product strategy. You'll be responsible for ensuring our AI capabilities are best-in-class while enabling our product teams to ship AI-powered features that delight users.</p> <h1>Teams You'll Partner with</h1> <h2>AI Research</h2> <p>The AI Research team leverages, trains, and validates powerful models for our product use cases across two core areas:</p> <ul> <li><strong>Audio/Video Research</strong>: Models for understanding, augmenting, and generating audio/video content (transcription, lipsync, video regenerate, TTS, avatars, etc.).</li> <li><strong>LLM Research</strong>: Evaluating and optimizing LLMs for Descript products, co-designing agent architecture, experimenting with token optimizations and fine-tuning.</li> </ul> <h2>AI Enablement</h2> <p>The AI Enablement team supports integrating 1P and 3P models into the Descript product:</p> <ul> <li>Building and maintaining standardized 3P model integrations (LLM providers, generative model APIs).</li> <li>Productionizing 1P models for specific use-cases.</li> <li>MLOps infrastructure (evals framework, inference infra, training infra, data pipelines).</li> </ul> <h1>What You'll Do</h1> <h3>Strategic Prioritization</h3> <ul> <li><strong>Make build vs. buy decisions</strong>: Evaluate when to train our own models vs. integrate third-party solutions based on market gaps, competitive advantage, and ROI</li> <li><strong>Balance research investment</strong>: Allocate team resources between long-term research bets, feature work, and maintenance</li> <li><strong>Guide research direction</strong>: Use product insight to inform what the team trains and develops; use research understanding to guide product direction</li> </ul> <h3>Evals & Quality</h3> <ul> <li><strong>Own the evals strategy</strong>: Design evaluation frameworks that are productionized and tied to real user needs, not just academic metrics</li> <li><strong>Drive quality standards</strong>: Establish quality bars for 1P and 3P models before they ship to users</li> <li><strong>Build feedback loops</strong>: Instrument data pipelines to continuously learn from user behavior and improve model performance</li> </ul> <h3>Cross-Functional Orchestration</h3> <ul> <li><strong>Partner with product teams</strong>: Advise on which models or architectures are best suited for specific features over time</li> <li><strong>Enable fast iteration</strong>: Build infrastructure and processes that let product teams experiment with AI capabilities quickly</li> <li><strong>Manage dependencies</strong>: Coordinate research timelines with product roadmaps and feature launches</li> </ul> <h3>Cost & Infrastructure</h3> <ul> <li><strong>Optimize COGS</strong>: Make strategic decisions on model selection, caching strategies, and infrastructure to balance quality, latency, and cost</li> <li><strong>Scale research infrastructure</strong>: Ensure the team has the DevEx, training infra, and tooling to move fast</li> </ul> <h2>Required Experience</h2> <p><strong>Product Sense</strong></p> <ul> <li>4+ years of product management experience, with at least 1-2 years working on AI/ML products</li> <li>Track record of making sound build vs. buy decisions in the AI space</li> <li>Experience balancing research exploration with shipping product value</li> <li>Ability to translate technical capabilities into user-facing product features</li> </ul> <p><strong>Technical Foundation</strong></p> <ul> <li>Understanding of modern ML/AI systems and LLMs (you don't need to write the code, but you need to understand the tradeoffs)</li> <li>Experience shipping AI/ML products to production at scale</li> <li>Experience with evals frameworks, model training pipelines, and inference infrastructure</li> <li>Understanding of ML cost structures (training compute, inference costs, token economics)</li> </ul> <p><strong>Cross-Functional Leadership</strong></p> <ul> <li>Experience working with research teams and helping them focus on high-impact work</li> <li>Track record of partnering with engineering teams on infrastructure and platform…