Twitxy Saved

Emma

Verified account This account is verified. Learn more

@EmmaQian_

CEO @Nova_ai - we're hiring! Forbes 30 Under 30. Ex Google DeepMind.
New York, NY novaintelligence.com Joined April 2021
561 Following 1.9K Followers 93 Posts

Latest posts 5

Emma Verified account @EmmaQian_
We’re hosting a happy hour this Thursday (Oct 15) in NYC, bringing together top engineers from OpenAI, DeepMind, Jane Street, Jump Trading, and more.

Drinks, food, poker, and good company at @nova_ai’s HQ.

A few spots left — reach out if you’re interested.
71 replies 13 reposts 305 likes 34K views View on X
Emma Verified account @EmmaQian_
Come join us for dinner on Sep 17 in NYC — a great group of engineers from Anthropic, DeepMind, Jane Street and more. A few spots left, reach out if you want in!
alexa Verified account @AlexaKayman Sep 4
I'm hosting an invite-only dinner for talent at the frontier of AI, with engineers from Anthropic, DeepMind, Jane Street, + more.

Have a few spots left. Who else should I invite?

Join us for a italian-japanese fusion tasting menu and great drinks Sep 17 @ 7pm in NYC.

Co-hosted by @Accel, @chemistry, and @Nova_AI.
View quoted post on X
15 replies 3 reposts 67 likes 17K views View on X
Emma Verified account @EmmaQian_
Life update: I’ve started a company with my good friend @samuely4ng 🚀

We’re building vertical coding agents to transform a 100B/yr industry and just closed a $7M seed round from Tier 1 investors (more details coming soon!)

Now, we’re assembling a world-class founding engineering team in SF. If you or someone you know is interested in building novel agentic systems that can solve complex real-world engineering problems, please DM me or Sam — we’re offering a $10k referral bonus for successful hires 👀
61 replies 22 reposts 673 likes 85K views View on X
Emma Verified account @EmmaQian_
Ever wondered how much data you need for fine-tuning? In our experiments with OpenAI's finetuning API, we saw notable improvements in two tasks with ~100 data points. We also conducted cost and latency analyses and found surprising results (finetuned models are ~3.6x faster!)