Data Science @ OpenAI. Prev. @ Eppo, Stitch Fix, Stanford.
Joined November 2012
- Tweets 210
- Following 185
- Followers 255
- Likes 926
14 Photos and videos
Pinned Tweet
17 Oct 2022
How come Bayesian and frequentist guarantees can be so different, even when the results are the same? In 1994, Philip Dawid published a fantastic paper about this.
Let's look at this from an experimentation perspective!
schmit.github.io/statistics/…
1
3
4
Sven Schmit retweeted
8 Sep 2025
We are building Allyon: Battlecry, the first monster trainer where creatures actually learn from you. It's closer to the Pokemon anime than the games: Allyon understand your voice, learn bespoke behaviors, have personalities and memories, and form lasting bonds with you.
1/7
7
10
50
6,733
Sven Schmit retweeted
14 Dec 2024
In the January/February 2025 @HarvardBiz (and online now): a new article that explains the steps that companies should follow to get better at experimentation
hbr.org/2025/01/want-your-co…
from me, @IavorBojinov, @rameshjohari, @spschmit and Martin Tingley
2
1
726
Sven Schmit retweeted
20 Aug 2024
Eppo lands new cash to grow its app, website and AI experimentation business tcrn.ch/3X70jNf
5
15
22,092
Fuck it ship it.
I wrote this while kinda being annoyed by people asking for advice.
But know if that I sound angry and haughty, it’s only 40% intentional.
jxnl.github.io/blog/writing/…
53
46
668
140,330
Sven Schmit retweeted
22 Dec 2023
The last couple of months I've been working on a game called Thus Spoke Zaranova. A game where you as a human must pose as an AI. (1/7)
18
24
179
72,681
Sven Schmit retweeted
19 Jul 2023
We are now accepting submissions for @CODEConference 2023!!!
We welcome any and all research related to digital experimentation, whether it comes from academia, industry, or somewhere else. If you have any questions about whether your work is a good fit, shoot me a DM/email.
🚨Call for @CODEConference abstracts🚨
Submit your work by Sept. 15 for the chance to present @MIT at the 10th annual @CODEConference
🐦Early bird registration open now - Oct. 2
👉Submit your work and register here: bit.ly/mitcode23
7
26
10,575
When does it make sense to bandit, and when to experiment? Let @spschmit walk you through the decision-making process geteppo.com/blog/bandit-or-e…
2
1
1,112
7 Apr 2023
Fantastic talk, highly recommended!
6 Apr 2023
Last couple of weeks I gave a few talks on the Sparks paper, here is the MIT recording!
The talk doesn't do justice to all the insights we have in the paper itself. Neither talk nor twitter threads are a substitute for actual reading of the 155 pages :-)
youtube.com/watch?v=qbIk7-JP…
219
22 Nov 2022
A lot has happened this year in terms of ML/AI progress, but this one I personally find most exciting!
22 Nov 2022
New paper in Science today on playing the classic negotiation game "Diplomacy" at a human level, by connecting language models with strategic reasoning! Our agent engages in intense and lengthy dialogues to persuade other players to follow its plans. This was really hard! 1/5
19 Nov 2022
The day "statistically significant" goes from statistical jargon to empty buzzword(s)... Although, more likely I'm just (statistically) significantly late to the party.
19 Nov 2022
With 116.6 million followers, your polls are starting to become statistically significant. What if Twitter had an "All Users" poll that you could push to every single twitter account to find out what the entire network is thinking, with no particular adverse selection? 🤔
2
Sven Schmit retweeted
16 Nov 2022
Really amazing to see how the Stitch Fix styling algorithm has evolved. Very smart approach.
Client Time Series Model: a Multi-Target Recommender System based on Temporally-Masked Encoders multithreaded.stitchfix.com/… via @stitchfix_algo
1
6
Sven Schmit retweeted
9 Sep 2022
New article: Formulas for Bootstrapping Sample Medians
The applications are limited, but I had fun deriving the equations, and maybe you'll learn something.
Enjoy!
evanmiller.org/bootstrapping…
1
5
25 Oct 2022
It was great to see old friends and meet new ones at the @CODEConference last week!
Here's my personal list of favorite talks from the conference:
svenschmit.com/code-roundup
For those who attended, what were your highlights?
1
1
10
17 Oct 2022
How come Bayesian and frequentist guarantees can be so different, even when the results are the same? In 1994, Philip Dawid published a fantastic paper about this.
Let's look at this from an experimentation perspective!
schmit.github.io/statistics/…
1
3
4
17 Oct 2022
The crux of the paradox lies in the prior: Bayesian analysis is powerful, but requires a realistic prior to work well.
Don't expect any magic with an uninformative prior.
1
17 Oct 2022
Note, this is not a critique of Bayesian methods, but rather meant to add some nuance to an often heated discussion.
Neither Bayesian nor frequentist methods are inherently superior. They provide two different approaches that are difficult to compare; both have pros and cons.
1
Sven Schmit retweeted
18 Jul 2022
Been a while folks, but I've finally put up a blog post. CUPED is a technique for tricking data into letting you finish A/B tests early. I explain the underlying math – and propose an improvement.
evanmiller.org/you-cant-spel…
3
6
35