Live data from Hacker News

Viewing profile — Smerity

Smerity

HN member
Joined
Thu, Nov 19, 2009, 10:54 AM UTC
HN karma
5,243
Public activity
490 items

About Smerity

Australian in San Francisco focused on making the rapid deep learning advances usable by anyone.

gcc startup.c -o ./startup

I hold a Masters in Computational Science & Engineering from Harvard University and a Bachelor of Information Technology (First Class Honours + University Medal) from the University of Sydney.

Highlights:

- Deep learning: As a senior research scientist at Salesforce Research (via the MetaMind acquisition), I established multiple state-of-the-art results in deep learning research, primarily language modeling, beating results from Google, Facebook, and other strong research groups.

- Big data: Lone engineer at Common Crawl where I produced crawl archives comprising of over 34 billion URLs (~2.5 petabytes).

- STEM education: First employee of Grok Learning, an online interactive platform tailored to teach high school students programming, which teaches tens of thousands of students per year. Nearly spent a decade at the National Computer Science School in Australia where we teach 100+ high school students per year programming at an intensive summer school.

Say hi! =]

Email: smerity@smerity.com

Website: http://smerity.com

Like my comments? I share similar things on Twitter a few hundred characters at a time! https://twitter.com/Smerity

Recent public activity

  1. comment
    Comment #45284414

    Ton Roosendaal is brilliant both strategically (taking a closed source codebase and growing it into the open source success story that is Blender today) but also as a person. When …

  2. comment
    Comment #42563422

    You can download Common Crawl data for free using HTTPS with no credentials. If you don't store it (streamed processing or equivalent) and you have no cost for incoming data (which…

  3. story
  4. comment
    Comment #41736930

    Excited to see more people working on RNNs but wish their citations were better. In 2016 my team from Salesforce Research published our work on the Quasi-Recurrent Neural Network[1…

  5. comment
    Comment #38950210

    I think you've done a great explanation expansion except I believe it's ALiBi ("Attention with Linear Biases Enables Input Length Extrapolation"), a method of positional encoding (…

  6. comment
    Comment #38732251

    As mentioned it's trivial across the spread of compression algorithms supporting this type of behaviour (`gzip`, `zstandard`, `zip`, ...), the header in `zip` making it even more c…

  7. comment
    Comment #38731000

    In the distant past I was the lone engineer of Common Crawl almost a decade ago. Common Crawl heavily leverages the WARC format. My favorite capability of the WARC format borrows f…

  8. comment
    Comment #37671510

    That section of the text is ambiguous but it's $250k for 8 x H100s (PCI-e) and $350k for 8 x H100 (SXM). The prices above are plus or minus from Lambda Labs[1]. [1]: https://lambda…

  9. comment
    Comment #33636516

    To note, Figma use something inspired by but not quite CRDTs: > Figma isn't using true CRDTs though. CRDTs are designed for decentralized systems where there is no single central a…

  10. comment
    Comment #32637224

    I have had my XPS 13 laptop die four times in the past two years, needing a replacement logic board each time. The laptop is amazing when it's working, default Linux support too, b…

  11. comment
    Comment #32410678

    The author is focused on GPU texture compression for games[1] so they're not too concerned with fast compression speed. Given a game will be played on (at a minimum) tens of thousa…

  12. story
  13. comment
    Comment #31940741

    I saw a version of this myself at Geary and Mason around midnight on June 20th[1]. That it happened again but at a larger scale ten days later is deeply concerning. Given I had no …

  14. story
  15. comment
    Comment #30955762

    I just want to note the replies to this thread are excessively dismissive and toxic. You may not agree with the wording of their advertising ("world's most powerful NLP toolkit" is…

  16. story
  17. story
  18. comment
    Comment #30299038

    What I find most interesting in this is the pseudo detective story of hunting down disappearing post-mortem and "lessons learned" documentation. Optimistically we'd hope that perha…

  19. story
  20. comment
    Comment #28779673

    From the pricing page I linked to[1]: > There are no charges for outbound data transfer when using a dedicated network connection to the RSTOR network and direct peering is establi…

  21. comment
    Comment #28778184

    You're right, I misread it and now can't edit my comment. This puts it all in quite a similar position to Wasabi[1] who have a 1:1 storage:transfer ratio per month and a 90 day min…

  22. comment
    Comment #28775942

    Thanks! I'd never seen them. The only minor catches that I've seen are that they want you to start the account with $50 or more added, it's $6 per TB per month (rounded up to the n…

  23. comment
    Comment #27232731

    The state of Australia's national science initiatives and the projects they support have always been of concern to me. There are bright spots but they seem overwhelmingly destined …

  24. comment
    Comment #26298026

    I loved Mia.bet's pet hamster marble race concept but their website notes: > The continuously high Ethereum gas fees cause betting fees and operation costs to skyrocket. Hence, we …

  25. story