Adi Kulkarni

41 posts

Adi Kulkarni

Adi Kulkarni

@AdiKulkarni6

CEO @ David Data | YC S26 | CS & Econ @ Yale

Katılım Nisan 2020
64 Takip Edilen76 Takipçiler
Alex Sima
Alex Sima@sima_alexx·
@AdiKulkarni6 Lmfaooo from now on when I'm late to a function, I'm going to start saying I'm almost done thinking.
English
1
0
0
30
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
"almost done thinking" is the new "5 minutes away"
Adi Kulkarni tweet media
English
1
0
5
256
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
do ppl still use bloomberg terminal? feels like its falling off cuz of all the new financial data providers
English
0
0
3
96
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
agents setting up emails, api keys, accounts… they’re probs gonna be doing YC in the fall at this rate
English
0
1
7
981
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
anyone’s customer support inboxes getting flooded by users agents instead of actual ppl? trying to figure out how common this is
English
0
0
4
718
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
The model sees what a real analyst would on the day (prices, financials, earnings, estimates, news, filings), it’s point in time. For the prices, we don't simulate actual buyers and sellers. It's a model tuned to move like a real market (calm periods, volatility spikes, crashes clustering, reactions to earnings/macro), and events like earnings or shocks push the prices around.
English
0
0
0
11
David J Yang
David J Yang@djyang1·
@davidhfdata whatchu know about my data jk but can you speak the the observable processes that were supposed to be indexed on for this RL problem? i.e. behind the market how did you model demand?
English
1
0
1
71
David Data
David Data@davidhfdata·
Frontier LLMs are impressive at many things. Trading is not one of them. We had Fable, GPT-5.5, Gemini 3.5 Pro, and Grok 4.5 each run a portfolio for ~1.6 years on a synthetic stock market they've never seen. All of them lost money. Think you can beat the benchmark? Try it 👇
David Data tweet media
English
8
2
29
4.1K
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
Testing on real market data after the model’s knowledge cutoff could be a more accurate representation of real-world accuracy, but yeah the time window is much smaller and is only one path, while synthetic data lets us test across a much wider range of market conditions (and we validate the synthetic data on real panels).
English
1
0
2
24
Bence Redmond
Bence Redmond@BenceRedmond·
@davidhfdata really interesting! is it possible to simulate the model's performance since the knowledge cut-off instead of simulating it? the time frame is a lot smaller but may be a more accurate representation of real-world accuraccy? or is the synthetic data equivalent?
English
1
0
3
117
Alex Sima
Alex Sima@sima_alexx·
Taking the leap is hard, but there's joy in the unknown. @AdiKulkarni6 and I spent senior year pivoting through 4 ideas (including an AI hedge fund). Now we're building @davidhfdata — synthetic market data for superhuman AI trading. Excited to join @ycombinator S26!
Alex Sima tweet media
English
3
6
30
9.5K
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
Superhuman AI traders won't be trained on the past. They'll be trained on every possible future. That's what @sima_alexx and I are building at @davidhfdata. Excited to join @ycombinator S26
Adi Kulkarni tweet media
English
9
1
26
620
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
@ojaskandy In one world our model created, Goliath wins and David is cooked
English
0
0
2
28
Thijs
Thijs@cdngdev·
surprisingly, 99% of sf tech workers have never crossed this line
Thijs tweet media
English
131
18
949
145.2K
Adi Kulkarni
Adi Kulkarni@AdiKulkarni6·
15 TB for a Minecraft world? that's a war crime on grass blocks.
English
1
0
1
131
Alex Sima
Alex Sima@sima_alexx·
emperor claude got me feeling like anakin vibewalker in revenge of the sith @sama
Alex Sima tweet media
GIF
English
1
0
2
195