Modulate is hiring a Senior Machine Learning Engineer to work on voice AI problems across research and engineering to design, train, evaluate, and deploy ML models powering the conversational voice intelligence platform.

Modulate is a frontier AI lab building the voice intelligence layer for the internet. We develop foundation models for audio: models that don't just transcribe speech, but understand it: who's speaking, what's being said, whether it's real, and whether it's safe. Velma is Modulate's audio-native voice intelligence model, purpose-built for audio from the ground up — not a text model retrofitted for speech. Velma powers a suite of APIs for developers building the next generation of voice-first products: → Modulate Transcribe: state-of-the-art speech-to-text. #1 on the Hugging Face Open ASR Leaderboard (out of 88 models), with top results on Sierra's μ-Bench, including #1 for Mandarin (zh-CN) → Deepfake Detect: real-time detection of AI-generated and cloned voices → AI Music Detection: identifying AI-generated audio content at scale → PII/PHI Redaction: automated sensitive-data protection for regulated industries Developers use Velma to build voice products that are faster, more accurate, and safer by default — without stitching together brittle, single-purpose audio tools. ToxMod, our proactive voice moderation system, applies this same model foundation to real-time safety — trusted by studios including Ubisoft, Activision, and Schell Games to detect toxicity and harassment without relying on player reports. Modulate was founded on a simple premise: audio is its own modality, and it deserves models built natively for it - not adapted from something else.

Founded 2019