deepa (@deepakorani0) header
deepa (@deepakorani0) profile picture

Nerd | Consistently brainstorming

Jersey City, NJ
Joined July 2021
Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)Photo by deepa (@deepakorani0)
Why I hate Instagram A women showing skin has more followers than a women with a PhD blog. Next generation is definitely going to be messed up
73
61
22
760
Given Claude Fable 5.2 can predict protein sequence, I was curious if it can predict structure. I gave it a relatively simple prompt: "predict contacts, 2ndary struct, use dynamic prog. and linear alg. to predict structure" It was able to predict fold of ubiquitin correctly 🤯
Claude was trained on protein sequences, at scale. We gave 20 language models the start of little-studied bacterial proteins and asked for the next amino acids. Only the latest Claude models predict them. Not GPT, not Gemini, not seven open-weight models.
33
62
13
413
42,653
Took a decade to switch lanes 😀
109
Biotech founders in Boston/NYC, let’s talk. I’m meeting people building in the space this month. DM me and I’ll send a link.
68
Given investors trust LLM judgements more these days…here is opus5.5 response on the slop 🤣.
We built an 8000 sqft. high-throughput wet-lab in 30 days, to solve molecular discovery - starting with pesticides. Most teams in drug-discovery don’t receive target-organism verification until clinical trials. Pesticides allow us to rapidly validate our model in the target organism directly. With the help of our high-throughput lab, we were able to generate compounds that killed a $13 billion pest, the Fall Armyworm. We believe that the bitter lesson is coming for biology.
187
Introducing LiteFold Dynamo. For the first time, you can now run 100ns+ molecular dynamics simulations at LiteFold. Molecular dynamics is one of the last steps before researchers validate in wet labs. At LiteFold, run multiple simulations, manage them, get insights, and view trajectories in real time within the platform. Check it out, it's free. Links in the comments.
8
17
2
118
12,485
Very soon.
1
6
3
56
4,605
Week 1 of @theresidency @enter_delta is nearly over. Modern drug discovery needs a team of infrastructure, AI/software engineers, computational biologists, medicinal chemists, and biologics experts. What if one platform combined them all, letting scientists focus on solving complex, novel problems? That's what we're building at LiteFold. Check out LiteFold, link in the comments—it's free.
18
12
2
135
7,022
Read ProteinCrow from future house and the first thought : kills most startups building workflow orchestrations in comp biology space 😀
98
What a interesting read
This paper claims LLMs are better at selecting successful founders than VCs "We introduce VCBench, the first benchmark for predicting founder success in venture capital (VC)" "most models surpass human benchmarks"
2
140
“Action produces information. Just keep doing stuff.” — Brian Armstrong
29
462
31
3,471
158,705
Are you applying to @ycombinator Fall batch? Repost this tweet and I’ll review your application to help boost your chances At @fluently_app, we were part of the W24 batch, and I applied twice before getting in. So I can provide valuable feedback for you for free Offer valid only next 24 hours ⏳
5
29
1
82
12,523
An O-1 visa is the best choice for any high potential US immigrant We recorded a 2-hour course with @LoganUllyot showing you exactly how to qualify & apply The course is $299 - but for the next 24 hours, it's free. To get it, - Like & RT this post - Comment ✈️ & I'll DM you
1,115
714
12
1,606
412,208
The last few months, I’ve been helping develop the mission for @FutureHouseSF - a non-profit AI Biology institute. Building on our work with autonomous agent driven chemistry and robotics, we’re building an entire org around the mission to build autonomous scientific agents 1/5
21
65
5
499
122,953
"agents-with-tool-use"
26
22
6
324
114,541
📢📢 New release @ProjectTDC - Clinical drug development 💊🏥 New tasks & datasets with 17,000 clinical trials that include validated trial outcomes and other data, including disease and drug information, and trial eligibility criteria Thanks @TianfanFu tdcommons.ai/multi_pred_task…
4
13
1,879
Now you can use AutoTrain Advanced to finetune any LLM available on Hugging Face Hub on any custom dataset locally! Soon, you will also be able to do training on Hugging Face Spaces. Try it now: pip install autotrain-advanced! 🚀
3
48
2
247
47,334
How YouTube got started:
13
102
4
659
121,239
deepa retweeted
The version high-skilled immigration that will be acceptable to everyone: —Stop H-1B fraud —Add a point-based system which accounts for English skill, education and work level —Eliminate country caps on EB green cards in favour of 2 —Allow entrepreneurs
53
47
7
637
111,994
GPT-4's model is FINALLY out: it's 8 experts of 220B param models! When Google build GLaM, a 64 experts of 64B param, in 2021, it beat 175B param GPT-3 on many generation and understanding tasks. Perfectly plausible that GPT-4 annihilated GPT-3 with this architecture! 🧵 1/3
3
14
134
50,793