Siyuan Gao

AI @ Linkedin. PhD @ Yale. Yahh Yahh Yahh!

[Yi Lu @yyyiiillluuu]

What if your agents got better every time they ran?

Today we're launching Reflexio — the self-improvement platform for AI agents. It learns from live production traffic, so every interaction makes your agent smarter.

Right now, failures, corrections, and wins in production get thrown away — and improving task success from those signals is slow and manual. We felt this pain firsthand.

Reflexio closes the loop: when an agent fails, it captures why, distills it into a playbook, and the next run — for any user — starts with that lesson learned.

The results so far:
• 36% fewer task failures across customer deployments
• 47% better responses in head-to-head comparisons
• 57% lower token cost on GDPval when a similar task has been seen before

And it compounds: playbooks keep refining themselves with every new interaction.

Sign up free at https://t.co/zJjKxnIxpl — first 30 days of Pro on us.

[Stephanie Noble (@sneuroble.bsky.social) @sNeuroble]

NOA in hand, we are ready to launch research in the NeuroPRISM lab!🚀 Appreciate @noahghola @ResearchAtNU @NUGlobalNews taking time to profile this @NIMHgov R00-funded project

Now hiring a fully-funded postdoc—reach out if interested & spread the word!

https://t.co/CEiqiZqCQC

[Dustin Scheinost @DScheinost]

A bit late in posting this …

Now out (and on the cover!) of the June issue of AJP: “Transdiagnostic Connectome-Based Prediction of Craving“

Using task-based functional connectomes, we predicted self-reported cravings in 274 individuals with CPM.

https://t.co/h1wovxAOl9

[Javid Dadashkarimi @JDadashkarimi]

In new work out in Medical Image Analysis, we present  CAROT: Cross Atlas Remapping via Optimal Transport. CAROT touches on some of the vibrant areas of functional MRI, predictive modeling, neuroimaging, & connectomics including privacy & open science 1/16 https://t.co/mVhmCZarfB

[Stephanie Noble (@sneuroble.bsky.social) @sNeuroble]

Thrilled to announce @psychonetrics & I are joining the @Northeastern Center for Cognitive & Brain Health as TT Assistant Professors THIS SUMMER!🎉🥳🍾I'll be in Psych+BioEng!

We were struck by the sense of community in the fast-growing Center—a perfect home for our new labs
1/3

[An Wu @SearchForAnWu]

#MontréalFire Day 5 after An's missing
Her parents' Canada Visa was approved yesterday. High living costs in Canada & US + travel will be a big burden on the family’s financial situation. We are raising funds to help cover their travel/living expenses: https://t.co/vsu9NUdHF5

[Stephanie Noble (@sneuroble.bsky.social) @sNeuroble]

Do you do:
👶 fetal/infant/toddler (FIT) neuroimaging?
💻 machine learning?

Want to do both?

If so, we’re excited to announce a new primer and review on FIT-specific machine learning
—a reflection on 100+ studies from the past decade
https://t.co/uVAMP1v3vL
1/5

[Siyuan Gao @CarrRickyyy]

@XiaoxiaoLi8 @yifannnwu oh wow! Looking forward to the photos!

[Abigail Greene @abigails_greene]

1/ So excited and proud to share our latest work, out now in @Nature: https://t.co/VVpKGfplcV. Despite all we’ve learned from them, models relating brain activity to phenotype fail consistently in a subset of people—specifically, people who defy sample stereotypes. A brief 🧵:

[Siyuan Gao @CarrRickyyy]

@M_Farruggia @DScheinost @danamsmall Congrats!!!

[OHBM Open Science Special Interest Group (OS-SIG) @OhbmOpen]

For our 5⃣ #OSR panel, Dustin Scheinost @DScheinost & Roza Gunes Bayrak @redgreenblues will lead a discussion on Social Bias in Machine Learning. June 23 (10:30 GMT+1) - Featuring four terrific panelists:

[Stephanie Noble (@sneuroble.bsky.social) @sNeuroble]

Can’t be there in person, but I’m there in spirit! #OHBM2022

We previously showed how broader-scale/multivariate inference improves power. Our poster builds on that by showing how much we then sacrifice specificity
+@DScheinost @mandyfmejia @AndrewZalesky

Feel free to DM any qs

[Link Tejavibulya @LinkTejavibulya]

Grateful to have worked with such a fantastic team including
@MaxRolison @CarrRickyyy @liang_qinghao Hannah Peterson @JDadashkarimi @M_Farruggia Alice Hahn @sNeuroble @lichsar @AngelPollatou @ajdneuro @DScheinost 2/

[Siyuan Gao @CarrRickyyy]

@ajdneuro Congrats, Alex!

[Gal Mishne @gmishne]

We're hiring postdocs @HDSIUCSD and jointly in collaboration with the Simons Institute at Berkeley
https://t.co/UCtnU0hOPp
https://t.co/8TFzYaDtxu
Please RT!
#postdoc #postdocjobs #datascience #AcademicJobs

[Max Rolison, MD @MaxRolison]

Great work led by @LinkTejavibulya Large-scale differences in functional organization of left- and right-handed individuals using whole-brain, data-driven analysis of connectivity @abigails_greene @CarrRickyyy @sNeuroble @DScheinost https://t.co/Sd5o6mliKn

[Kangjoo Lee, PhD @kangjoolee.bsky.social @Kangjoo_Lee_]

Thrilled to announce our latest work! Here, we use simultaneous resting state fMRI and pupillometry to study the impact of arousal levels indexed by pupil area on the integration of large-scale brain networks. 1/n https://t.co/o76hdYkKJe

[Todd Constable (find me here @toddc.bsky.social) @constable_todd]

1/3 Excited about this new work by @WendyLuo_ demonstrating loads of functional connectivity information contained within-nodes of a typical atlas.
https://t.co/zmfc5ggeq5

[Javid Dadashkarimi @JDadashkarimi]

I am excited to share my latest work about optimal transport with @DScheinost and @aminkarbasi :

Title: Data-driven mapping between functional
connectomes using optimal transport

preprint: https://t.co/0XCXi1FKpp

code: https://t.co/ntPjLXCJlK

#MICCAI2021 @MICCAI_Society

[Siyuan Gao @CarrRickyyy]

6) Link to code
https://t.co/FiEdVy8Ak0

[Siyuan Gao @CarrRickyyy]

Excited to see this paper with @gmishne and @DScheinost published! We embedded both the task and resting-state scans from #humanconnectomeproject on a low-dimensional temporal #manifold. More details in the thread
https://t.co/nZmHhlSWpp

[Siyuan Gao @CarrRickyyy]

5) There are more interesting results in the paper: comparison to PCA, embeddings relationship with participation coefficient, etc. Check out the paper if you are interested!