Dev (@DevvMandal) header
Dev (@DevvMandal) profile picture

nerd | computer-use data at @markov_labs (YC S26) | ex @SarvamAI | IITM ‘27

sf
Joined January 2024
Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)Photo by Dev (@DevvMandal)
Today, we're announcing Markov (YC S26). We've built the most advanced CUA datasets for CAD, design, browser-use & more. We're working with frontier AI labs and our open-source datasets have crossed 150,000+ downloads on HuggingFace. Sample dataset below :) @markov__ai
83
28
12
597
59,018
Today, we're releasing SolidWorks-Bench: can AI models do real mechanical engineering tasks in SolidWorks? Models get close on individual parts but struggle to assemble them. No model completes a single machine. Fable 5.2 leads with a 57% mean reward. Full report below
46
23
13
463
54,236
if you'd like to know more about our engineering/CUA datasets and environments, book a time with me here: cal.com/dev-markov/30min
1
6
2,620
Maya-2 voice model weights are officially live! 🇮🇳🇮🇳 The frontier voice model for India is now completely open-weight. • Outperforms ElevenLabs v3, Sarvam Bulbul v3, and Cartesia Sonic 3.5 • 1.5B+ people covered across all 28 states and 8 union territories • SOTA voice synthesis across 11 official languages • 20 native speakers pre-loaded, ready out of the box • Best-in-class zero-shot voice cloning Built by a team of five from the land of languages. Go build and have fun!
199
354
45
3,435
142,695
Dev retweeted
From building robots to running a college. On April 14th I took over as CEO of Regional College of Management Bangalore. Why? The institution had huge potential, the challenges were fascinating, and honestly, I love working with our students. They come from all over the country, mostly tier 2 & 3 cities. Brilliant kids. They just need polish. Turns out the job isn't that different from building robots. It's system design: for ops, for how students develop, for innovation. And a lot of it runs on AI, from day-to-day productivity to technical systems to media. My path doesn't make much sense on paper. Robots, then a college, then who knows. But it's always been the same thing: find hard problems and build systems to solve them.
1
4
218
We’re releasing the world’s best Indic voice AI model, fully open-weight. 🇮🇳🇮🇳 With over 1.5 billion people and 22 official languages, the future of AI for India is voice-first. Maya-2 officially outperforms ElevenLabs v3, Cartesia Sonic 3.5, and Bulbul v3 across Indian languages. It will be completely free for every builder very soon.
185
314
25
2,582
303,175
introducing kyra glasses. a proactive wearable that handles your communications and work. it figures out what needs you, handles what it can, and brings you in when it genuinely matters. today, your phone makes you manage everything yourself. kyra is built to handle the rest. pre-orders open now.
221
90
90
669
356,985
YC has been an absolute blast and I'm so excited to present our incredible progress at demo day! (if only my co-founder could have made it too 😔)
Kind of bummed I'm not in SF for @ycombinator demo day, but I'm proud of what we've built at @markov_labs these last three months. Started the batch when I was 17, turned 18 last week, feels like a whole chapter went by. Cool stuff coming soon, can't wait to share!
9
2
61
4,274
Update: Samarth and I have joined OpenAI. We started Kairos with a dream that felt too big to say out loud: a personal AGI for everyone. Excited to keep chasing that dream inside Codex and ChatGPT.
246
33
20
2,137
502,311
1/ Introducing Sodium Sodium turns all your website’s existing features into WebMCP tools so AI agents can discover and use your site directly. > 2 minute setup > 2.82x faster > 91x token efficient sodium.result.dev
85
89
20
1,292
129,727
Another insane open-source model just dropped!
Introducing Phonon-1: a new frontier in speech recognition per byte. A 782 million param model in a 415 MB download, more accurate than a Whisper 2x its size. Transcribe an hour of audio in about two minutes on a MacBook Air. Open weights, Apache 2.0, today. @fermion_ai
1
23
3,778
i can now sit in front of ms paint and yap until there’s an actual painting. made this with @alma_inc. try it out: alma.inc/paint
41
25
23
271
90,933
Today, we're launching the world's largest open-source CAD dataset for computer-use. 1,000+ hours across SolidWorks, Siemens-NX, AutoCAD & more. Screen recordings, mouse/keyboard actions, input diagrams, output files, rubrics and narrations. Link in the comments :)
43
148
29
1,414
160,985
This is just a small portion of what we have. If you'd like to know more about our proprietary datasets, book a meeting with us here: cal.com/dev-markov/30min?ove…
1
10
2,735
Tomorrow we're launching the largest open-source dataset of computer-use data for CAD. Across Siemens-NX, Solidworks, AutoCAD and more. Including task prompts, input files, screen recordings, synced mouse/keyboard actions and output files. Stay tuned :)
Today, our computer-use datasets crossed 50,000+ hours. Screen recordings + synced mouse/keyboard actions + narrations across CAD, design, browser-use & more. Sourced from humans and available off the shelf today :) Samples below.
34
105
4
1,179
79,166
we've been moving at an incredible speed, roughly 6 years worth of continuous computer-use data collected.
Today, our computer-use datasets crossed 50,000+ hours. Screen recordings + synced mouse/keyboard actions + narrations across CAD, design, browser-use & more. Sourced from humans and available off the shelf today :) Samples below.
2
1
40
1,783
Today, our computer-use datasets crossed 50,000+ hours. Screen recordings + synced mouse/keyboard actions + narrations across CAD, design, browser-use & more. Sourced from humans and available off the shelf today :) Samples below.
15
15
3
237
84,126