Program Manager (Data Operations)

Salary: $175K - $220K Employment Type: Full-Time Location Type: Remote

Company Overview

Deepgram is the leading voice AI platform for developers building speech-to-text (STT), text-to-speech (TTS), and full speech-to-speech (STS) offerings. Over 200,000 developers build with Deepgram’s voice-native foundational models, accessed through APIs or as self-managed software, due to unmatched accuracy, latency, and pricing. Customers include software companies building voice products, co-sell partners working with large enterprises, and enterprises solving internal voice AI use cases.

Deepgram ended 2024 cash-flow positive with 400+ enterprise customers, 3.3x annual usage growth across the past 4 years, over 50,000 years of audio processed, and over 1 trillion words transcribed. The company states, "There is no organization in the world that understands voice better than Deepgram."

At Deepgram, data is not just fuel for their models; it's a product in its own right. Their vertically integrated voice AI platform depends on strategically created, curated, and labeled audio, from original data collection and augmentation to structured workflows and evaluation sets. These data pipelines support a range of cutting-edge technologies, including audio intelligence models, conversational voice agents, STT, and TTS.

Position Overview

Deepgram is seeking a hands-on, systems-minded Program Manager (Data Operations) to lead the design and execution of various voice data programs. This role is ideal for individuals who thrive on building from scratch, can take an abstract modeling goal or product need and turn it into a concrete data strategy and pipeline with tools, guidelines, and quality safeguards in place. This requires ownership to understand frontier research strategies, build custom style guides, prototype new tools, and directly influence how data shapes their products.

This is a role for builders who can spot a gap, roll up their sleeves, and design the solution. You will be at the center of Deepgram’s model development cycle, working across Research, Engineering, and Product, and will be deeply involved in both the day-to-day execution and the systems thinking required to scale it.

Responsibilities

• Design, launch, and own end-to-end data workflows: from raw audio ingestion to production-ready datasets.
• Build and evolve labeling specs, style guides, and instructional documentation for global annotation teams.
• Identify opportunities for better tooling, automation, and workflow optimization, and lead their implementation.
• Translate product goals and model requirements into data creation strategies, deciding what to build, how to build it, and why it matters for product impact.
• Prototype and deploy data tools and infrastructure (e.g., Label Studio, custom Python scripts).
• Collaborate with Research and Engineering to align data collection with model training architecture and downstream product impact.
• Track advancements in speech AI research and evolving market use cases to inform labeling approaches and data design priorities.
• Partner with QA and Evaluation leads to deliver high-quality, human-in-the-loop datasets and benchmarks.
• Manage and mentor data vendors, freelancers, and potentially internal ICs as the team grows.
• Track throughput, data quality, and vendor performance.
• Drive continuous improvement in speed, cost-efficiency, and quality across all data operations.
• Curate and refine datasets to align with specific product goals, linguistic coverage, or research hypotheses.

Ideal Candidate Profile

You’ll love this role if you:

• Believe that data is a product, not just a resource, and want to help define what great voice data looks like.
• Enjoy turning experimental ideas into robust, repeatable systems that can scale to production.
• Like to build systems that sit at the intersection of model performance, product impact, and operational excellence.
• Thrive in ambiguity and take initiative without waiting for perfect specs.
• Prefer action over perfection and iteration over indecision.
• Are energized by working with voice and speech technologies.