hanrelan
Joined 13 years ago
Karma
69
Submissions
52
Member since
Sep 21
Recent submissions
Comment · 1 month ago
Comment · 2 months ago
How do you deal with auth/bot detection/captchas in that browser? That's where my agent consistently gets stuck
Comment · 11 months ago
If you run a pentest, allowing rooted devices will almost certainly show up as a vulnerability. It'll be marked "low risk", but you'll also be told that you don't want to "accept risk" for too many "low risk" vulnerabilities.
So somebody then needs to say that this is not something they worry about rather than doing the easy thing and remediating it.
Comment · 1 year ago
ScribbleVet (https://scribblevet.com) | Full-stack, former technical founders | Full-time | Remote (must be able to work US timezones)
ScribbleVet is an AI-powered scribe for veterinarians. We save veterinarians two hours every day, reduce veterinary burnout and help animals get better care.
* We're generating real ($XM/yr) revenue
* Our users love the product
* We're fully remote but not asynchronous, so we have a lot of autonomy but it's easy to collaborate with someone when needed.
We're looking for:
* Senior engineers (former founders or early startup experience strongly preferred) who can work across the stack.
* Engineers with a product mindset. Customer-centricity or design skills are a plus.
Our tech stack: Postgres, Elixir (+ Ash framework), React Native/Expo, React/Nextjs
We’ve found that the engineers who are happy and successful at Scribble really enjoy technology (learning new frameworks, experimenting with cutting-edge platforms) and are very product-oriented.
If you're interested in building AI-driven features that help vets and the animals they care for, check us out here: https://excessive-grey-9af.notion.site/Senior-Full-Stack-Fro...
Comment · 1 year ago
ScribbleVet (https://scribblevet.com) | Designer (marketing & product) | Full-time | Remote (must be able to work US timezones)
ScribbleVet is an AI-powered scribe for veterinarians. We save veterinarians two hours every day, reduce veterinary burnout and help animals get better care.
- We're generating real ($XM/yr) revenue
- Our users love the product
- We're fully remote but not asynchronous, so we have a lot of autonomy but it's easy to collaborate with someone when needed.
We're looking for:
- A designer to join our team (full-time) to work across product design (UX) and marketing assets (landing pages, trade show materials etc)
- Strong product skills and founding, early employee, or building something from scratch experience required
If you're interested and think you might be a fit, please contact me at rohan (at) scribblevet.com!
Comment · 2 years ago
You'll probably find this talk [1] interesting. They control all the training data for small LLMs and then perform experiments (including reasoning experiments).
[1] Physics of LLMs: https://www.youtube.com/watch?v=yBL7J0kgldU&t=7s
Comment · 2 years ago
ScribbleVet (https://scribblevet.com) | Full-stack (frontend emphasis), former technical founders, AI engineers | Full-time | Remote
(must be able to work US timezones)
ScribbleVet is an AI-powered scribe for veterinarians. We save veterinarians two hours every day, reduce veterinary burnout and help animals get better care.
- We're generating real ($XM/yr) revenue
- Our users love the product
- We're fully remote but not asynchronous, so we have a lot of autonomy but it's easy to collaborate with someone when needed.
We're looking for:
- Senior engineers (former founders or early startup experience preferred) who can work across the stack and are AI curious
- Engineers with a product mindset. Customer-centricity or design skills are a plus.
Our tech stack: Postgres, Elixir (+ Ash framework), React Native/Expo, React/Nextjs
If you're interested in building AI-driven features that help vets and the animals they care for, check us out here: https://www.notion.so/Senior-Full-Stack-Frontend-Engineer-6b...
Comment · 2 years ago
Any videos in particular you'd recommend?
Comment · 2 years ago
Is Modal a good solution for running fine-tuned LLMs and Whisper models? If the cold-start time is low we're more than willing to modify our code to use Modal's infra.
Happy to follow up via email but didn't see one in your profile.
Comment · 3 years ago
It requires API access, but once you have access you can easily play around with it in the openai playground.
Setting temperature to 0 makes the output deterministic, though in my experiments it's still highly sensitive to the inputs. What I mean by that is while yes, for the exact same input you get the exact same output, it's also true that you can change one or two words (that may not change the meaning in any way) and get a different output.
Comment · 3 years ago
You should be able to get the data pretty easily from opendota: https://www.opendota.com
Comment · 3 years ago
I think you can go far quite cheaply. Get your code working on smaller/toy models, and then when you want to test it on larger ones you can ship it over to a machine at one of the cheaper providers (vast.ai/jarvislabs etc) to give it a run before pausing/killing the machine.
I've been porting Stable Diffusion (which isn't a small model) over to Elixir and as part of doing that have been starting/stopping my jarvislabs machine when I start/stop building. I've been spending about $1/day without trying to be efficient.
Also, fast.ai is a great resource for learning ML, I highly recommend it.
Comment · 3 years ago
I've trained models using Jupyter and Livebook (though only smaller toy models [1]) so I can deposit my 2 cents here. Small disclaimer that I started with Jupyter, so in some sense my mental model was biased towards Jupyter.
I think the biggest difference that'll trip you up coming from Jupyter is that Livebook enforces linear execution. You can't arbitrarily run cells in any order like you can in Jupyter - if you change an earlier cell all the subsequent cells have to be run in order. The only deviation from this is branches which allow you to capture the state at a certain point and create a new flow from there on. There's a section in [1] that explains how branching works and how you can use it when training models.
The other difference is that if you do something that crashes in a cell, you'll lose the state of the entire branch and have to rerun from the beginning of the branch. Iirc if you stop a long running cell, that forces a rerun as well. That can also be painful when running training loops that run for a while, but there are some pretty neat workarounds you can do using Kino. Using those workarounds does break the reproducibility guarantees though.
Personally while building NN models I find that I prefer the Jupyter execution model because for NNs, rerunning cells can be really time-consuming. Being able to quickly change some variables and run a cell out of order helps while I'm exploring/experimenting.
Two things I love about Livebook though are 1) the file format makes version control super easy and 2) Kino allows for real interactivity in the notebook in a way that's much harder to do in Jupyter. So in Livebook you can easily create live updating charts, images etc that show training progress or have other kinds of interactivity.
If you're interested to see what my model training workflow looks like with Livebook (and I have no idea if it's the best workflow!), check out the examples below [1][2]. Overall I'd say it definitely works well, you just have to shift your mental model a bit if you're coming from Jupyter. If I were doing something where rerunning cells wasn't expensive I would probably prefer the Livebook model.
[1] https://github.com/elixir-nx/axon/blob/main/notebooks/genera... [2] https://github.com/elixir-nx/axon/blob/main/notebooks/genera...
Is it possible to flash them from Google home firmware to esphome without USB? I might move away from Google home soon but I'd hate to have to reopen up my ACs again.