Modulate, a Boston voice intelligence company, has raised $25M to push audio-native models into more enterprise deployments. Future Ventures led the round, with Hyperplane and Lakestar participating.
The company runs more than 100 small models split across two jobs: extracting signals such as emotion, tone, language and whether a voice is synthetic, and analysing intent, for example whether a caller is breaking rules or attempting a scam. Because the models stay small, they do not demand specialised accelerators, which helps as token bills climb.
Founded in 2017 by MIT physics classmates Mike Pappas and Carter Huffman, Modulate began with voice modulation for gaming, moved into voice moderation, and now sells deepfake detection, agent quality scoring and compliance monitoring for regulated industries.
Politeness is not satisfaction
Huffman argues transcription alone misses the nuance that decides whether a call went well. Customers may stay polite while deeply dissatisfied, so a simple positive-or-negative read misleads. Modulate sits beside a company’s existing voice stack and grades both human agents and AI agents.
PitchBook put Modulate’s prior funding at $41M and its last valuation at $170M. The startup has 40 to 45 staff, plans to add about 10 more, and is building on-premises and on-device options for privacy-sensitive buyers.
