Vivek Jayaram
@vivjay30
Founder @VocalityHealth (YC W25) | PhD in AI/ML | Prev founder @fulltrackai (4m users) | Sharing insights on building AI products and startup journeys.
Excited to share our new paper: Constrained Diffusion Implicit Models! We use diffusion models to solve noisy inverse problems like inpainting, sparse-recovery, and colorization. 10-50x faster than previous methods! Paper: arxiv.org/pdf/2411.00359 Demo: huggingface.co/spaces/vivjay3…
Unpopular opinion, if you can find people in India to do quality work for your company, you absolutely should. Costs 1/5th as much as US engineers, meaning you can provide a livelihood to 5x as many people, and do 5x the work with the same budget.
REAL-TIME INFO FROM INSIDE IBM VIA *CONFIRMED* INSIDER: "IBM laid off 9k US workers this year and replaced them with “AI” -- when in reality they’ve had over 3k new people join in their India offices. Nearly 85% of senior leadership is Indian and they keep pushing for jobs…
I’m trying to learn how to promote B2C apps on TikTok. Anyone have any good tutorials or guides? AI influencers or pay for human influencers?
The Windsurf situation shows that it’s never been a better time to be a founder and a worse time to be an employee: Founders get stock, employees get options at ridiculous FMV with 30 days to exercise. Founders are selling secondaries earlier than ever (series A or 😱 even at…
"you don’t seem to understand, i have a PhD in ML, i was meant to pretrain language model" "wrap the fucking API"
The clearest sign that we haven’t reached AGI is that AI can’t autonomously make money
New Anthropic Research: Project Vend. We had Claude run a small shop in our office lunchroom. Here’s how it went.
One of my favorite intuitions about gradient descent is that local minima are exponentially less common in high dimensions. That’s why neural nets don’t get stuck during training.
In machine learning, we take gradient descent for granted. We rarely question why it works. What's usually told is the mountain-climbing analogue: to find the valley, step towards the steepest descent. But why does this work so well? Read on.
In the long run, the only software companies that will exist will be foundational model research companies and undifferentiated sales shops.
Love to see this new open source music generator from Google!
Excited to announce 🎵Magenta RealTime, the first open weights music generation model capable of real-time audio generation with real-time control. 👋 **Try Magenta RT on Colab TPUs**: colab.research.google.com/github/magenta… 👀 Blog post: g.co/magenta/rt 🧵 below
I remember when publishing several first author papers in top conferences was the bar to finishing a PhD, not starting one.
Ok a few quick things. Most CS students who get into elite PhD programs in AI especially have already usually published multiple first author papers in "top" conferences. 1/
Graduated from my PhD at UW today A huge thanks to my advisors, labmates, friends, family, reviewer #2, the Twitterverse, the lab espresso room, and everyone else who helped me along the way!
