@MIT alum | Formerly @NASA @NASAJPL | Current CEO & Co-founder of @Modulate_ai
Joined August 2015
- Tweets87
- Following84
- Followers35
- Likes42
Pinned Tweet
I used to teach spacecraft to see.
Now I teach software to listen.
@modulate_ai recently raised $25M, led by Future Ventures, to build the listening layer for AI.
Cosmology → JPL → voice. Here’s the path 🧵
Starting a new series: Velma in the Wild.
We take famous scenes, run Velma on the raw audio, and watch what it hears. No script. No transcript. Just the sound.
Episode 1: Adolescence @netflix
Watch Velma catch anger, amusement, and the exact moment amusement turns into frustration, straight from the voices.
Words give you the script. Voice gives you the performance.
Working in voice AI ruins small talk. In the best way.
At @modulate_ai, we listen for what words alone can't tell you. 🎙️
Heading to SF next week for the Masters of Scale Summit, my first as CEO of @modulate_ai
New role, same question: how do you build AI people can actually trust?
This year’s theme is human agency as AI gets more capable. Right up my alley.
Thanks @kaliouby for the invite.
Say hi if you’re there.
#MastersOfScaleSummit
Our deepfake detection: #1 on the Hugging Face deepfake leaderboard. 98.9% accuracy on synthetic audio.
Cloned voices keep getting better.
So do we.
Words are only 3–20% of what a voice conversation communicates. The rest is in the tone, pace, and emotion that transcripts miss.
That’s an architecture problem, not a model quality problem.
I broke down why at Ai4 this year, including why 59%+ of consumers report frustration with AI voice agents.
Watch the full talk: bit.ly/3SWlwua
Then put audio-native AI to the test. Get 1,000 free Modulate API credits: bit.ly/4gMSTZt
Your next customer call could be a deepfake.
The usual fix is more verification steps. Slower calls. Angrier customers.
Don and I are going live on LinkedIn Oct 20 to dig into the trade-off: how do you stop voice fraud without wrecking the customer experience?
Spoiler: the answer isn’t in the transcript 👀
What’s the most frustrating thing a voice assistant has ever done to you?
I’ll start: kept talking for eight full seconds after I said “wait.”
The future of voice AI already aired. On Star Trek.
You talk. The computer just gets it.
No rephrasing. No “no, I said set a TIMER.”
Today, you adapt to voice AI.
In a couple of years, it adapts to you: hears frustration, reacts like a person would.
Building voice AI is one thing.
Knowing if it works, and if you can trust it, is another.
On two panels at VON today for @modulate_ai:
- 10:40 AM, Main Stage: VoiceAI Rising Stars
- 3:50 PM: How Do You Know Your Voice AI Actually Works?
Spoiler: the transcript isn’t enough 👀
Sharing the stage with @jeffpulver, VoiceRun, @AgentPhoneHQ, Relay Hawk, @webrtc, @Vobiz_Ai and Grandstream. Come say hi.
vonevolution.com/f26-oct13-v…
Building voice agents is half the problem.
Knowing what’s actually happening in those conversations is the other half.
In Atlanta for VON, demoing @modulate_ai at VoiceAI LIVE! today alongside @SoundHound and others.
2:40 PM, Room C. Come say hi. I’ll be listening to how you say it.
