Log inSign up
Dan Biderman
Engram
2,427 posts
user avatar
Dan Biderman
Engram
@dan_biderman
CEO @EngramLab Previously @HazyResearch @cu_neurotheory @DbrxMosaicAI
Palo Alto, CA
dan-biderman.netlify.app
Joined March 2017
1,563
Following
3,376
Followers

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • Pinned
    user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Jun 23
    The single biggest impediment in AI isn't absolute intelligence or coding ability. It's deeply understanding large repositories of knowledge that every person and every company has. Our repositories will explode in size, with AI agents writing much of them. This is a two-way
    user avatar
    Engram
    @EngramLab
    Jun 23
    Article cover image
    Article
    Introducing Engram: Scaling compute on your context
    We’re Engram. We’re building AI that learns from you and deeply understands your work. Today’s AI models don’t understand what you do. Not really. Everything models know comes from their training –...
    34K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Feb 25, 2025
    How can we use small LLMs to shift more AI workloads onto our laptops and phones? In our paper and open-source code, we pair on-device LLMs (@ollama) with frontier LLMs in the cloud (@OpenAI, @Together), to solve token-intensive workloads on your 💻 at 17.5% of the cloud cost
    00:00
    193K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    May 17, 2024
    People think LoRA is a magic bullet for LLMs. Is it? Does it deliver the same quality as full finetuning but on consumer GPUs? Though LoRA has the advantage of a lower memory footprint, we find that it often substantially underperforms full finetuning. However, it forgets less
    173K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Jul 8, 2025
    Do you believe in local LLMs?
    215K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Jul 26, 2023
    LLMs are hard to evaluate, and evals are hard to visualize. To make our lives easier, we built a little interactive @streamlit + @plotlygraphs app that plots flexible eval metrics, such as "world knowledge" and "common-sense reasoning" from @MosaicML's Model Gauntlet.
    GIF
    26K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Jul 20, 2023
    ⚡️Announcing️Lightning Pose⚡️ A pose estimation method that learns from both labeled images and a lot of unlabeled videos, via inductive biases like temporal smoothness, posture feasibility, and multi-view consistency. Preprint: tinyurl.com/litpose
    27K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Aug 13, 2024
    *LoRA Learns Less and Forgets Less* is now out in its definitive edition in TMLR🚀 Checkout the latest numbers fresh from the @DbrxMosaicAI oven 👨‍🍳
    user avatar
    Certified papers at TMLR
    @TmlrCert
    Aug 13, 2024
    New #FeaturedCertification: LoRA Learns Less and Forgets Less Dan Biderman, Jacob Portes, Jose Javier Gonzalez Ortiz et al. openreview.net/forum?id=aloEr… #learns #regularization #rank
    36K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Aug 9, 2021
    A standard way to accelerate Gaussian Processes is to tradeoff computation for bias. In our ICML paper we quantify the biases of Random Fourier Features and Conjugate Gradient algos. We then instead tradeoff computation for variance, to get better results proceedings.mlr.press/v139/potapczyn…
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Jun 26, 2024
    ✨Paper out in final form: exciting results from our semi-supervised pose estimation package, Lightning Pose, which is now adopted by a number of great neuroscience labs. Please give it a whirl: github.com/danbider/light…
    user avatar
    Nature Methods
    @naturemethods
    Jun 26, 2024
    Lightning Pose is an efficient pose estimation approach that requires few labeled training data owing to its semi-supervised learning strategy and ensembling. @dan_biderman @cu_neurotheory @ZuckermanBrain @IntlBrainLab @Columbia nature.com/articles/s4159…
    9.5K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Aug 18, 2025
    Pretraining might be back. DeepSeek/Kimi/etc are telling us how to build the models Datology/etc tell us how to curate the data Not impossible for small teams to own decent models
    6.7K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Mar 12, 2025
    apropos the Gemma release @GoogleDeepMind and @ollama
    6.5K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    May 23, 2025
    Part of my speech at the Columbia Medical Center PhD hooding ceremony: “The world we’re graduating into now has machines that review literature, write code, analyze data, and even draft manuscripts that pass peer review. Our scientific profession is shifting beneath our feet.
    7.7K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Mar 2, 2025
    Local LLMs are powerful enough to clean user prompts from private information before sending to cloud APIs. We're in iter=0. Help us make LLM calls 100% private. @Avanika15 💪 @ollama cc
    user avatar
    Avanika Narayan
    @Avanika15
    Mar 2, 2025
    [3/1] minions daily ship 🚢 🔒 privacy mode ft. papillon (@Sylvia_Sparkle, @databricks) and @spacy_io turn on privacy mode — only your minions see your private information 🙈
    00:00
    12K
  • user avatar
    Dan Biderman
    Engram
    @dan_biderman
    Feb 25, 2025
    Replying to @dan_biderman
    With co-first authors @Avanika15, @EyubogluSabri at @HazyResearch. Paper: arxiv.org/abs/2502.15964 Blog: hazyresearch.stanford.edu/blog/2025-02-2… Ollama Blog: ollama.com/blog/minions GitHub: github.com/HazyResearch/m… NotebookLM Podcast: tinyurl.com/4v9b3rwc
    10K