Hadar

I do AI

[Hadar @SharvitHadar]

Excited to be presenting our work at #ICLR 2025!

If you’re attending, come find me during Poster Session 4 (Today 3:00–5:30 PM SGT, Hall 3 & Hall 2B),
poster #43:
“A Deep Inverse Dynamics Model for a Flapping Robotic Wing.”

[Barak Raveh @BarakRaveh]

Presenting DeepDynamics, a joint effort by the @naomi_habib and @BarakRaveh Labs. We have built a deep learning framework that bridges bulk and single-cell RNA-seq to map the cellular cascades underlying Alzheimer’s Disease (AD) and aging. #alzheimer #deeplearning #ROSMAP
1/10

[Hadar @SharvitHadar]

Which team would you join?

[Hadar @SharvitHadar]

chasing SOTA neural net architectures is mostly a cool distraction.

Consider two teams:

[Hadar @SharvitHadar]

Team 2️⃣
Uses an “okay” model. Built a solid data pipeline. Constructed a loss function that encodes the actual business goals. Deployment is scalable, observable, and designed for fast iteration across all assets.

[Guy Yariv @guy_yariv]

We present DyPE, a framework for ultra high resolution image generation.
DyPE adjusts positional embeddings to evolve dynamically with the spectral progression of diffusion.
This lets pre-trained DiTs create images with 16M+ pixels without retraining or extra inference cost.
🧵👇

[Hadar @SharvitHadar]

@guy_yariv Sounds super interesting. I’m a fan of transformation that are sensitive to frequencies

[Hadar @SharvitHadar]

Check out my latest article: Nested Tensors https://t.co/75qY8w9MTg via @LinkedIn

[Hadar @SharvitHadar]

🌍 Digital nomads, we built this for you.
WiFi Atlas is a live 3D map of real WiFi speeds, powered by travelers like you.
📡 Real-time speed tests
🧳 Crowd-powered signal map

No more guessing if the café WiFi sucks.
Drop a 🌐 in the replies to get early access + updates.

[Hadar @SharvitHadar]

First thoughts on #Kiro
-Unique planning with design/tasks/requirements markdowns, similar to realworld planning
-Generators tests for each implementation: these usually do not tackle core functionality, and miss the point
-managed to beat copilot/cursor on some
-free
100% try it

[Hadar @SharvitHadar]

Playing with 3D globes in JS is fun. What am I building? 😅

[Hadar @SharvitHadar]

can't say it's much, but I'm investing 0 minutes/month on this app, and it still brings in $2/mo 😅
thank you, one user

[Hadar @SharvitHadar]

New doom scrolling platform: #GitHub
Just spent 1 hour skimming through cool GitHub repos.
immediately hooked.

[Hadar @SharvitHadar]

@robj3d3 what are you making?

[Hadar @SharvitHadar]

@weswinder especially love step 3.

[Hadar @SharvitHadar]

I launched https://t.co/j87Ejbe6k3 a few months ago, but it never got traction.
Fast forward a few months: two influencer‑backed clones are live and working beautifully.
lesson learned: B2C lives on social. I just needed a sharper social‑first marketing play.

[Hadar @SharvitHadar]

@dumay_sacha I just listened to a podcast about a couple who built a virtual try-on API for small businesses. It works on a token-based payment model, allowing you to upload customer photos and visualize them wearing your actual clothing items.

[Hadar @SharvitHadar]

while the Saas solo-founder stories pile up, I'm starting to hear more and more about solopreneurs that serve an API, SDK, etc. Wonder if this is a valid approach in 2025

[Hadar @SharvitHadar]

https://t.co/ndjPU7xuhj

[Hadar @SharvitHadar]

As data scientists, it’s easy to forget the domain we’re dealing with once the data is big enough and the models complex enough. But when we do that, we miss some of the most important aspects of machine learning, and honestly, we lose some of the fun.