Product Manager, Text-to-speech

About Gradium

Gradium is a voice AI company building the Text-to-Speech (TTS) engine for the next generation of voice agents.

We believe the best voice agents will win on quality, and that quality starts with the voice itself. We're purpose-built for product engineers and AI-native companies who need voices that sound genuinely human, stay reliable in production, and don't break on real-world content like acronyms, alphanumerics, or cross-lingual names.

We are a small team, early but moving fast, with real customers and a clear thesis: voice agents are the next major frontier of applied AI, and the TTS engine underneath will matter enormously. Our goal is to become the default voice layer for voice agent applications, starting by matching the quality bar, then outrunning the competition on speed of iteration and developer trust.

What you're stepping into: small team, unreasonable ambition, high ownership, fast decisions, and very little process unless it earns its place.

The role

We are looking for a Product Manager to own the quality and reliability of Gradium's TTS models in production. This is a technically demanding role at the intersection of applied ML, evaluation infrastructure, and developer trust. It is the role that makes a Gradium voice not just sound great in a demo, but hold up on real-world content, on every release, in every language we support.

Your mission is clear: make our models "bulletproof" in production. Productize our TTS evaluation pipeline so every model iteration is measured before it ships, eliminate the pronunciation and hallucination failures that block adoption, and ship at least one new model that matches ElevenLabs on pronunciation accuracy across our core EU languages (FR, ES, PT, DE), validated with real customers. You'll work hand in hand with Thomas (CPO), research, and engineering to define what "production-ready" means and enforce it before anything reaches customers.

What You'll Do

• Run our evals and own the internal app: Own our internal eval app, with robust data engineering and async offline audio generation, able to run experiments at scale across every criterion, speaker, and supported language. Ship automated reports and dashboards for every model iteration so we can see progress and catch regressions before release, not after.

• Understand customer problems and arm the sales team: Get deep on what customers actually struggle with, from pronunciation failures and hallucinations to the edge cases that block adoption in production.Turn that into prioritized problems for research, and support the sales team with the demos, evidence, and quality answers that move deals forward.

• Drive the day-to-day of model iteration: Maintain a tight feedback loop between customers and research, and set the tempo for prioritization, model cycles, and releases. Keep research and engineering aligned on what ships next and why, and use levers like rewriting and self-distillation rules alongside model improvements to close gaps fast.

• Communicate our quality, inside and out: Publish a live benchmarking dashboard comparing models on robustness, expressivity, naturalness, and latency. Turn eval results into honest benchmarks and "wow" demos, and contribute to marketing content and publications that build developer trust and put our quality in front of the people who buy.

• Spread learnings across the product: Feed what you learn from evals and customers into other models and product areas, from docs and developer experience to Studio. Generalize the evaluation methodology toward ASR and voice selection, objectifying quality where it is subjective today, and help standardize training and testing data engineering with research.

• Work in an AI-native way: Use AI-assisted tools for prototyping, analysis, and evaluation, from automated quality scoring to vibe coding. Push the team to adopt AI-native workflows wherever they accelerate speed or improve output.

Who You Are

• Generative AI / ML product experience: You have shipped at least one product powered by generative AI or ML. You understand model evaluation, output quality variance, and the unique PM challenges of non-deterministic systems. You know the difference between a demo and a production-grade AI feature.

• Founder mindset: You act with urgency, take full ownership, and don't wait for permission or perfect information. You are comfortable making highstakes decisions in ambiguous environments and see the founding team as partners, not hierarchy.

• Obsessive about quality at scale: You've built systems or processes that maintain a high quality bar while increasing throughput. You think in terms of pipelines, automation, and measurable standards, not just subjective review. You've dealt with the tension between speed and craft and found ways to have both.

• Technically fluent: You can hold your own with research and engineering on model behavior, data, and infrastructure. You understand what an API is and who developers are, and you can reason about latency, accuracy, and reliability trade-offs.

• AI-fluent operator: You actively use AI-powered tools in your daily work, from vibe coding and AI-assisted development to automated analysis and prototyping. You see these workflows as table stakes for a modern PM, not optional extras.

• Structured and data-driven: You bring rigor to ambiguous problems. You define metrics, build dashboards, and use data to prioritize, but you also know when to trust your judgment and move fast. You've operated in earlystage environments where you had to build the measurement infrastructure yourself.