Paavas Bhasin

🚨 Just dropped: Perplexity unveils a built-in language learning feature! Thanks to founder Aravind Srinivas and his brilliant team! 🗣️ Whether you want to brush up on Spanish before vacation or master Mandarin for business, you can now do it right inside Perplexity with interactive flashcards, AI-generated travel phrases, and instant translations. Check out the demo video below to see it in action! What language will you try first? #LanguageLearning #EdTech #Perplexity #ProductLaunch #LearningRevolution

10

2 Comments

Excited to share an interesting paper by our intern — a refine-and-judge pipeline that iteratively curates data for LLM fine-tuning. The approach has shown promising results on benchmarks including AlpacaEval and MT-Bench.

20

1 Comment

I’ve been loving the new work coming out of Stanford’s Digital Economy Lab and The Budget Lab at Yale -- not because it shows dramatic change, but because it’s finally asking the right questions. The most honest conclusion we have right now about AI and labor is that we simply need more (and better) measurement and a readiness to support workers even before we have definitive answers. Going to spend the next few weeks sharing more here on my read on the work. Early shifts in AI-exposed jobs, especially at the entry level, are not proof that firms are “hiring AI instead of people.” Higher exposure predicts jobs where AI could plausibly reshape how firms think about junior work: 1/ jobs where most learning can happen off the job 2/ jobs where junior and senior work/tasks look pretty similar, with speed and polish as the main differences. Those are precisely the jobs where training data already exists -- where models can be trained if labs have access to the right data. So if those jobs are changing -- in number, scope, or quality -- it could reflect many forces. It might be delayed hiring, re-sequencing of skill development, or a rethinking of what early career work even is. It does not automatically follow that senior workers have suddenly become vastly more productive thanks to AI. This (correlation != causation) matters because it has real implications for how we design policy responses. If the symptom is that early career workers are struggling, then the right interventions should be tied to those workers -- new pathways, better training models, wage protections, smoother transitions -- rather than tied narrowly to AI -- subsidizing adoption, enforcing impact assessments, or chasing productivity metrics. If we assume the wrong mechanism, we will build the wrong response. The goal isn’t to respond to "AI." It’s to respond to what’s happening to workers. Bharat Chandar, Erik Brynjolfsson, Ruyu Chen, Martha Gimbel, Molly Kinder, Joshua Kendall, Madeline Lee

243

8 Comments

Recently I gave a talk at SF Ruby about a library we built (and recently open-sourced!) at AngelList called Zaxcel. Why'd we build it? Excel is the lingua franca of finance, but most programmatic Excel generation libraries force you to think in coordinates and formula strings. We took a different approach: model the workbook as a directed graph where data flows through named objects. Let the library handle coordinate resolution and formula composition. The result is a DSL that actually makes sense: you write formulas as Ruby expressions, reference things by name, and the type checker catches mistakes at compile time. It's revolutionized how we build financial statements. What used to take weeks now takes hours (and with AI agents - only minutes!), empowering thousands of workbook deliveries a quarter. Read how we went from nightmare refactors to 100x productivity: https://lnkd.in/es2aTttR Open-sourced repo (you can use it today!): https://lnkd.in/ecKRwcaW

91

8 Comments

On Friday, May 23, EchoHer hosted a Cofounder Matching workshop at Homebrew for early-stage founders. Dar Patel of Little Pursuits and startup mentor Bing Li shared insights on how to know if you need a cofounder, what makes a strong match, red flags to watch for, and lessons from real founder experiences. Video by: Emanuel Francis https://lnkd.in/gmd-hCZC

11

2 Comments

What if you could go from ideation to chip fabrication in seconds, with just a prompt? That’s the idea behind ChipChat, the app that my awesome teammates (Aditi Jain, Alka Ukrani, Yashna Hasija) and I built in 24 hours at the UC Berkeley AI Hackathon, winning 1 of 3 grand prizes! From just a prompt (like “build a 4-bit adder”) or even an image of a hand-drawn circuit, ChipChat generates: • Verilog code to define the behavior of the circuit • an interactive circuit schematic to virtually test the circuit • a chip layout file (DEF) that declares the physical locations of gates and connections • a GDSII file that can be sent directly to a chip manufacturer for fabrication ChipChat can be incredibly useful to help students gain a better understanding of digital circuits. And with hundreds of billions of dollars being put into chip development each year, we envision tools like ChipChat being used to save significantly on both money and time in the future. A huge thanks to the Cal Hacks team for organizing this hackathon! Had a ton of fun hacking and exploring the University of California, Berkeley campus. For how we built it + more details, check out our Devpost: https://lnkd.in/g3egwxVu

174

29 Comments

$200M for One Mind: Meta’s Bold AI Talent Bet As a CS undergrad, I’ve been closely following the evolution of AI - not just the technology, but the talent behind it. And this week, one headline truly stood out: Meta has reportedly offered over $200M to recruit Ruoming Pang, Apple’s former head of AI models. Pang isn’t just any engineer—he led Apple’s 100-person AI modeling team and spent 15 years at Google before that. His move to Meta marks more than just a career change—it signals the intensity of the AI talent war now raging across the tech world. What does this mean for the future? - Meta is clearly betting big on becoming the leader in superintelligent AI—systems that push the boundaries beyond human capabilities. - Compensation for AI talent is skyrocketing, resetting industry expectations for what it takes to attract top-tier minds. - More than ever, innovation is becoming a race not just for algorithms, but for the people who build them. As a student passionate about AI, this moment feels both inspiring and eye-opening. The landscape is shifting fast, and behind every model is a team of minds being fiercely competed over. My takeaway? In the age of superintelligence, the true superpower may just be talent. #AI #Meta #Superintelligence #ComputerScience #TechTalent #FutureOfWork #AIcareers #RuomingPang #CSUndergrad #MachineLearning #ArtificialIntelligence https://lnkd.in/g_8bW246

1

15th hackathon win 🏆 Snow Leopard AI Hackathon by Deepti Srivastava and Moraa O. The problem we kept thinking about? Commuters safety in San Francisco. People say New York is bad. SF is worse. In NYC, unsafe areas are spread out - you can route around them. In SF, encampments and incidents are concentrated in few streets in the heart of the city. You can't avoid them. You have to walk past them to get to work, to commute, to live. Built SafeWalk SF with Sissi Wang - an AI agent that pulls real SFCity data and visualizes it on a map, so you can see patterns street by street.  Going around in SF? Ask:  - What's the safest route from SOMA to Chinatown?  - Is it safe to park near Tenderloin for 4hours?  - Which streets should I avoid in the Mission? The agent queries real SF safety records (crimes, traffic injuries, encampments, etc) using Snow Leopard AI, writing SQL at runtime instead of maintaining a complex data extraction layer. Visualizes incident patterns on an interactive map Generates multiple walking routes - optimized for safety, speed, or scenery The outcome? You can see exactly which streets have repeated incidents, recognize the patterns, and make informed decisions - not based on fear or rumors, but on actual data.

50

5 Comments