Explainer
Who Made Dessa AI? Meet the Team Behind Block's Speech Model
Most AI launches are credited to a company. Dessa's is credited, unusually precisely, to a handful of named people. Block announced Dessa on October 23, 2026 — a full-duplex, speech-only voice model trained from scratch on human conversation — and its official page names the lead, the researchers and even the interns behind it. This article answers the question readers keep arriving with, in every phrasing from “dessa by block” to “who made Dessa AI”: who actually built this thing, where they came from, and what Block has and has not said about them. For the model itself, start with what Dessa AI is; this piece is about the people.
The short answer: a very small lab inside Block
Block describes the makers in its own words: “We're a (very) small AI research lab at Block, founded by former members of Dessa, the Toronto deep learning startup.” The lab reports directly to Jack Dorsey, Block's co-founder. That reporting line matters. Dessa is not a feature team inside Square or Cash App shipping a voice button; it is a research lab sitting close to the top of the company, working on a problem — a model that learns to talk from sound alone — with no announced product attached to it. Block has announced no Dessa product, price, API or release date, and the team's own page frames the work as curiosity, better human–AI understanding, and safety, in that order.
The named team
The Dessa lead is Ragavan Thurairatnam, who has called Dessa's emergence “a GPT-3 moment for speech” — a claim we examined in detail in our look at the GPT-3 moment claim. Beneath the lead, Block's page names six researchers, in alphabetical order: Louay Hazami, Hashiam Kadhim, Qi (Alex) Liu, Rayhane Mama, Aydin Polat and Marc Tyndel. It names three interns: Cara Ji, Jibran Iqbal Shah and Dvir Zagury-Grynbaum. It even credits two former interns, Aryan Sharma and Olivia Yan. That is the whole public roster. Block has not published a total headcount, hiring plans, or biographies beyond the page's own summary — and anything beyond these names and roles would be invention, so we stop there.
Where the team came from: Toronto
The lab's founders came from Dessa, the Toronto startup founded in 2016 as DeepLearni.ng, which became famous in 2019 for a synthetic Joe Rogan voice — an early public demonstration of how convincing generated speech could be. Square acquired Dessa on February 7, 2020, and its people came with it. The full arc is in our history of Dessa, from Toronto startup to Block lab. On its current page the lab adds two lines of track record: it says its members have built deep learning models since 2012 and generative speech models for over eight years. In other words, the team that once showed the world a fake Rogan is now, inside Block, building the speech model Block is releasing most cautiously — safety researchers first, precisely because a voice this convincing cuts both ways.
Why the small-team detail matters
Scale is the usual story in AI: bigger teams, bigger clusters, bigger models. Dessa's public story is the opposite at every point. A tiny lab. Training on a single NVIDIA GB200 NVL72 system. Language learned, Block says, from hundreds of times fewer words than large language models need. Inference that runs locally on a MacBook Pro M4 Max or better. Whether those claims hold up is a separate question — Block has published no benchmarks, parameter count or technical paper, as our benchmarks explainer sets out — but the pattern is deliberate: Block is presenting Dessa as a research result from a small, senior group, not as a platform launch from a large product organisation. That also explains the rollout. A research lab publishes recordings, opens a waitlist, and asks outside safety researchers to find the risks it has not anticipated. It does not ship an API on day one.
What the team has not said
Just as important is what you will not find from the makers, however hard you search. No team member has announced a release date, a price, an API, model weights or a language list. No one has confirmed that any waitlist invites have gone out — our invite status check covers what is actually known there. The official page, re-read today, November 13, 2026, still ends the same way it did at launch: recordings, a waitlist, and an appeal to safety researchers, signed by the lead. If a credible Dessa announcement appears, it will almost certainly appear there first — not in a third-party explainer, and not in a social post promising access. Until then, the honest summary of who made Dessa AI is also the complete one: a very small Block lab with Toronto roots, six named researchers, three named interns, one named lead — and a model the rest of us can still only listen to.
Sources: Block's official Dessa page (block.xyz/dessa), re-read November 13, 2026; DessaAI Daily reporting linked above. DessaAI Daily is independent and unofficial, and is not affiliated with Block, Inc.
Frequently asked questions
Who created Dessa AI?
Dessa was created by a small AI research lab inside Block, founded by former members of Dessa, the Toronto deep learning startup Block's predecessor Square acquired in 2020. The lab reports directly to Jack Dorsey, and its lead is Ragavan Thurairatnam.
Who leads the Dessa team?
Ragavan Thurairatnam is named on Block's official Dessa page as the Dessa lead. He also signed the page's appeal for AI safety researchers to apply for early access.
How many people work on Dessa?
Block calls it a very small lab. Its official page names six researchers and three interns, plus two former interns. Block has not published a total headcount beyond the people it names.
🚀 Get Dessa launch alerts
Dessa is still a research preview. Join the free DessaAI Daily list and we will email you the day access opens — plus the week's biggest Dessa stories, once a week. No spam, no hype.
Form not loading? Subscribe on the signup page →