zo7

It's possible to train your breathing patterns, look up Buteyko breathing.

Awareness of breathing does not mean controlling your breathing, it just means noticing the sensations associated with it. Breathing can be incredibly pleasant!

  Location: SF Bay Area
  Remote: Yes
  Willing to relocate: No
  Technologies: Machine Learning, LLMs, Audio/DSP, Edge ML; Python, C/C++; PyTorch, TensorFlow, Lightning, ONNX, TensorRT, Pandas/Polars; Kubernetes, Dask, Prefect, Docker, AWS, GCP; ARM, Hexagon DSP, Nvidia Jetson  
  Résumé/CV: https://drive.proton.me/urls/CK4FED6FQW#vGllUr7m7VwW
  Email: flynn.hn@proton.me

Machine learning engineer with 8 years of startup experience building production systems for audio, robotic perception, and edge deployment. Specialties in deep learning for audio/DSP, distributed training infrastructure, and shipping models to embedded hardware.

Currently exploring LLMs/RAG/agents through side projects, and interested in more work in that space.

  Location: SF Bay Area
  Remote: Yes
  Willing to relocate: Yes to LA, NYC, or Berlin
  Technologies: Machine Learning, Audio/DSP, Edge ML; Python, C/C++; PyTorch, TensorFlow, Lightning, ONNX, TensorRT, Pandas/Polars; Kubernetes, Dask, Prefect, Docker, AWS, GCP; ARM, Hexagon DSP, Nvidia Jetson  
  Résumé/CV: https://drive.proton.me/urls/WWV395Z7YW#pwIJb0u3HxFs
  Email: flynn.hn@proton.me

Machine learning engineer with 8 years of startup experience building production systems for audio, robotic perception, and edge deployment. Specialties in deep learning for audio/DSP, distributed training infrastructure, and shipping models to embedded hardware.

Currently exploring LLM and agent tooling through side projects, and open to roles that combine that with my core background.

It's becoming more common in SF and LA too, although it's usually done by the promoter and not the venue.

There are startups that deliver food to other startups, printing revenue numbers using the same VC’s entire portfolio of companies, therefore sloshing around the exact same dollars in order to get an attractive revenue multiple and exit.

Wow, I hadn't realized that. I looked up the catering company one of my previous employers used and perhaps unsurprisingly they share investors.

Adding to this, there is a specific kind of meditation you could try called metta (aka loving-kindness) where you focus on cultivating benevolence towards yourself and all things. During a session you follow a similar sequence: first focus on cultivating loving-kindness towards yourself (since this is the base from where empathy grows) then you gradually extend it to people you love out to people who you have difficulty with. If mindfulness is an exercise to train your mind to observe itself, metta trains your mind to observe other minds.

Also second Rosenberg's book. It can come across as condescending if applied too heavily but it's a great analysis on language during conflict.

What about the processing in the veggie salami/bologna turns you off? Their meat counterparts are also heavily processed and are known to be carcinogenic.

Are these conflicting interests? Better bike infrastructure would mean that those who can bike would be off the road, which would mean less congested streets and highways for those who need to drive.

Commuting is also only one occasion you need transportation for. Many local errands (like grocery shopping) could be done on bike and would be cheaper than using your vehicle.

Be more iterative, experimental, and user driven

My heart has fallen at the thought of agile software development invading the creative process

I dropped out, started a few companies, and made enough to pay for my education.

Don't bother with GPA then, this sounds like something you could focus on instead. GPA is only relevant for new grads without any real experience.

The troubling aspect of it aside, a policy like that would result in a net negative for everyone else. Global industries rely on poverty to guarantee an abundance of low-paid (or not paid at all) workers. In their absence, they'd either need to be replaced by bringing wealthy people into poverty, or we'd have to brace for an economic upheaval as traditional industries are disrupted (which would likely also create poverty).

(b) There's one big difference between Facebook and generic advertising technology: the New York Times doesn't require you to sign up.

(c) Ad tracking knows which websites I visit. Facebook knows which nude parties I attended in 2005

Facebook tracks you off the site and has been known to build shadow profiles of non-users.

Does the hardware do anything interesting, or does it only process midi input and provide an interface to mix between generated sounds? From their instructions on creating new sounds, it looks like they're pre-computing the generated sounds: https://github.com/googlecreativelab/open-nsynth-super/tree/...

Given the number of plugs for Google products/projects/research (especially near the end) it's probably intended to be more of an ad for Google.

I feel like this is related to the information bottleneck idea that's been floating around for some time [1]. I only really understand both of these at a superficial level, but from what I think I understand one thing that they observed is that there are two phases when training a deep learning model: a phase which maximizes the mutual information (?) between the input and the output, and a compression phase which compresses the learned representation. In that light this work makes sense, since artifacts are essentially noise that the network would filter out in the fitting process.

Very cool work though.

[1] https://youtu.be/bLqJHjXihK8

The author mentions that they only analyzed the collection of C# repositories that they used in their study on Singletons, but I would expect the results to be different depending on the language used.

In my experience with Python, unit testing is essential to catch careless errors (typos, bad ducks, etc. that can't be caught by a linter) that would cause your program to crash at runtime, so there's a slightly different motivation for writing unit tests. It's a less restrictive language so you don't have the issues with private methods like the author mentioned, and is (at least to me) a bit less of a strain to write tests. Unit testing might correlate more with good hygiene for a Python codebase because testing is vital to check correctness and there are fewer situations where you have to mangle your code for the tests, whereas it may not in C# because you get guarantees from the type system that makes testing less vital and writing tests is more of a nuisance.

A far better way to write tests is within your production code. Validate every parameter and every assumption. Raise errors when possible, log everything else.

It runs when you testers, and customers use the applications.

Do you rely on any sort of automated testing to catch errors before it gets to your customers? From a business standpoint it doesn't feel right to offload testing to your users, especially in more critical systems.

I feel it's pretty arrogant to think that you can contribute to research entirely on your own. You really need to be around other researchers who are smarter and more experienced than you to share ideas with and learn from, you simply do not have the insight needed to make progress in your field on your own. If you were that wunderkind who could do it, you honestly wouldn't be asking.

If you truly want to get into research, grad school should sound like a dream (at any school with a decent program, not just top schools). You'd be surrounded by likeminded people, get to do crazy science stuff all day, and you'd push the frontier of knowledge a little bit. It's really the best and most effective way to immerse yourself in the material and become a good researcher.

You could strike out on your own only to find out there's not enough of a market to support yourself with your skills and you have to get a job again. You could set a goal to save money to travel the world and have it all be depleted recovering from a medical emergency that leaves you disabled. You could invest all of your money for retirement only to see it vanish in an economic crisis.

These are more realistic but there's still an element of chance that could wrestle control away from you. If you use goals to give purpose to your life, what will you do if you discover that you absolutely cannot meet them? If it was the source of your drive, will you be able to laugh it off? What if you get there and discover that meeting your goal was not something that you even wanted at all?

My point was that you could instead focus on things immediately under your control: you could learn skills and network which may help you become your own boss, but it may also help you get a better job or start a new career if you'd rather. You could save money which would allow you to possibly travel the world one day, or it might allow you to indulge in some other nice experience or simply help reduce stress in your life. You maintain control over your life and can seize opportunities as they come, but you're not constrained by stuff outside your control, or even your past self who was setting the goals to begin with.

I'd just be unhappy if I'd constantly be setting big life goals and not reaching them when planned, or at all.

This is key, I think. You need to have some acceptance that you actually have very little control over your life. Plan for and manage the parts of your life that you do have control over, and keep an open mind about the rest. Some of the unhappiest people I know are those with lofty goals that they either fail to meet, or have unrealistic expectations about what they will get for meeting them.

I feel like maintaining the infrastructure for the mammoth amount of data they must serve each day would be quite a lot of work on its own. They can't really afford hiccups in service, and I don't recall the site ever being down or a video not loading recently.

How so? In San Francisco there are several local produce markets within walking distance of where I live. Whole Foods and Trader Joe's are pretty common major grocery chains here, and there are two weekly farmer's markets I can think of that are accessible by bike/transit.

When I visited Manhattan/Brooklyn recently I didn't get the sense that they were lacking in locally-sourced foods either. Main problem I can think of are food deserts, but those are more of an economic problem than scale. When I used to live in one in Baltimore though it was possible to find local food, it just took more effort.

This seems like a plausible theory, especially since the article notes that online dating is overwhelmingly popular among homosexual couples, but doesn't that theory start to fall apart when a couple is going out together? Unless their relationship is entirely secret I'd imagine they'd be seen in public eventually (especially at a school) and they'd probably still experience hostility for it (and probably worse than just flirting).

I feel like this is still related to what the authors were claiming, that people previously only dated within their social connections. The schools I've been to were never overtly racist, but friend groups still seemed to contain mostly people of the same race/class. You'd have very few chances to meet someone outside your social group if your connections only consisted of your social group.

That was kinda my point, it's not the be-all deep learning library because they made it for their own use case, but its towering popularity (as in 10x the number of stars of other popular libraries) is not genuine.

Also I highly doubt that the main reason Google open sourced it was to be charitable.

I think the author raises a good point about Google envy. TensorFlow is not the most intuitive or flexible library out there, and it is very over-engineered if you're not doing large-scale distributed training. The main reason why everyone talks it up so much is because Google heavily marketed it from the outset, and everyone automatically assumes Google == Virtuoso Software Design because they couldn't make it through the interview. Really it's just modern enterprise software which has five different ways to implement batch norm that they push on the community so they don't have to train new hires on how to use it.

You know what? I started looking for more facts on this, and to be honest I don't think there are any. All of the news, especially since Trump tweeted this morning, is extremely politicized. For any source that provided evidence one way or another I could not untangle any objective facts from whatever agendas I felt they were trying to push. Most of what I found was just fluff that blew up some quip or minuscule fact that I honestly only thought was worthwhile because it was presented in a way to exploit my confirmation bias.

But this is the tactic this administration has been using to scatter people's attention and make it impossible to find reliable information about what is actually going on. Cause some outrage that exacerbates biases in the media, claim that news is unreliable because of bias, and then present the facts as you would like it.

Everyone realized the storm was worse than anyone predicted, mayor criticizes government and pleads for more help, DT creates some outrage, and now they can deflect any criticism by saying that their critics are out for their throats.

I have two undeniable facts though:

* Shit's bad.

* The WH's slow response and DT's comments are not helping.

There are a lot of charities that have been working to help areas affected by these recent storms and earthquakes. Is there an easy way to donate and have it be spread across several trusted charities at once? What is the best way to help here?

I didn't specifically reference Puerto Rico.

You commented on a post that was specifically about Puerto Rico.

Perhaps my frustration should be directed at leadership instead. I actually wrote that without any data on what the government has done vs what they could have done, but the white house at least has been notably slower to react now than with past storms:

https://www.washingtonpost.com/politics/lost-weekend-how-tru...