benkoller

--••• is indeed 7. Pretty neat part of the story.

This sounds like a great story, congrats! Having the opportunity to go "all in" on a hobby in this way is enticing for sure. Would you mind telling more about how the opportunity came up to go full-time? Was it being in the right place at the right time, or what type of risk did you step into?

Interesting approach, thanks for the good reads. Would love to support, no strings attached, if you see an opportunity. Reached out on LI, feel free to dump the request if this feels off to you.

Situational awareness is in fact a teachable/learnable skill and not just a "thing you have". Every professional field has one or another concept of the around it. I'd even go as far as saying it's one of my crucial (soft-)skills and has benefitted me greatly in my career, so it's great to see that I'm not alone with my interest in the topic.

As someone currently dabbling in the space, this is great reality checklist for a majority of research papers (and, unfortunately, a large majority on commercial endeavours, too) on the topic. The author makes a solid delivery on his outset promise.

It just might help to shed some light on something akin to the blockchain/machine learning hypecycle bubbles amongst startups.

In case someone finds it helpful, I created a terraform script to create all AWS resources to self-host a static site via S3 and Cloudfront: https://gist.github.com/benkoller/360019b4ad40d572bccc2875b4...

I've been using this for years at over a dozen different projects, I think it's a great way to host even commercial sites.

  Location: Munich, Germany  
  Remote: Yes
  Willing to relocate: No
  Technologies: Linux, Python, Kubernetes, all the things DevOps, MLOps
  Résumé/CV: https://www.linkedin.com/in/benediktkoller/ or https://benkoller.de/assets/CV-Benedikt_Koller.pdf
  Email: contact at benkoller dot de

---

I'm a DevOps guy at heart turned manager and founder of open-source MLOps tooling. I've scaled from 30 to 230, from 5 to 50 and most recently from 0 to 5. Along the way learned to love hiring/mentoring teams, building purpose-driven tech stacks, solving challenges data-driven and with a people-and-result-centered culture.

I wish you the best of luck and see a short-term need, but as someone deeply entangled into gastronomic subculture I would hate to see the experience of being hosted by a gastronomer and gastronomic team fade away. Going to your favourite restaurant or to a new, "fancy" place is just as much an experience in hospitality as it is in culinary.

Interesting to see that they relied on random forests for feature selection. I'd wonder if a more "classical" approach would have yielded similar/comparable results, or if these findings where only correlatable due to the use of ML.

Have you guys considered a open-core business approach? Would dramatically increase trust, and my willingness to throw money at you :)

This is a great approach to determine permission especially for service accounts - thanks for giving me another tool for my toolchain.

> Instead of creating a single modality i.e. artificial noses or ears, medicine would benefit from a multimodal sensory system that uses vision, smell, auditory and perhaps even taste, _in conjunction_, as the human body is rich with signals through all these modalities.

Thats a great thought right there. Adding an over-time perspective would make this even more powerful, as a lot of information is encapsulated in change, not just absolute observation.

I wonder how much of the required sensing tech is already available, but most likely not rated for medical usage yet.

Godmode troll activated.

The Matt Levine article is a great summary, thanks for sharing.

> If you too much shit to do, a calendar wont fix that. If your meeting are fucked because people are badly prepared, a calendar wont fix that. If you have to many appointments with your parents in law, your calendar wont fix that.

Couldn't agree more. People > Process, both ways around.

Any chance for a port to Big Sur? Would love to be #9.

Maybe it's me being blind, but I do not see a download/signup anywhere. How would one try bearicorn?

Great approach. I've been running ops teams for a long time, and eventually all teams have ended up spending some time optimising our CI tooling to involve some degree of caching - so you definitely are on to something.

Before I clicked your pricing I'd really wished for an affordable pricing plan to run this on BYO resources, but only your enterprise plan seems to cover this. I always get an iffy feeling when I have to build my software on external resources I have 0 control over. Your downtime will prevent teams from shipping their code - but I guess that thought is part of your upsell to the enterprise plan.

Anyway, it's one of these ideas I'd wish I had years ago, so congratz to you.

I'd be thrilled to hear how the energy consumption under Linux would compare to native Big Sur - obviously with the M1 Macbook Pros in mind.

Interesting - out of curiosity, why do you thing the TF ecosystem is dying?

ZenML unfortunately does not yet work with PyTorch, but it’s one of the biggest points for our roadmap next year.

Definitely! It might sound like a sales-thing to say, but PyTorch in general is a huge point on our roadmap, but we’re not 100% sure what the most logical approach is. To keep the APIs consistent will involve some fiddling on our end.

Hi, one of the co-founders here. We’ve been using the original base of ZenML for a few years in its previous shapes, mostly in predictive analytics projects. We obviously had to switch things up to make the code base suitable to a broader audience and are still working out kinks.

Especially feedback like yours is super appreciated, as we can gauge opinions and evaluate them against our roadmap, so thanks :).

Hi! We are fully bootstrapped, and ZenML is the final result of a longer journey. When we started over three years ago we were running predictive analytics projects with industry partners. Over time our focus expanded, until we finally realized: our tech stack is much more valuable. We did a lot of talking to people to find out if there is a niche for our perspective, and have been pushing the code base to where it is today for the last months.

So, to answer in short: there is a longer development cycle behind this, but not in the open - and no VC backing.

It really breaks down to acknowledging how many overlaps ML has with traditional software development. Great to hear that adding SREs worked for you - how did the transition go for the rest of the team? Any big learnings from your team about the transition?

I see your point. To me, the two points however don’t stand at odds, both need to be addressed just the same. I said something similar in other comments, but only if your organization is building a strong team structure AND a solid process/tooling structure projects in ML can go successfully into production.

To me these patterns feel very close to what organizations have been tackling for the last 20 years in software development. Teams need to be given the autonomy AND responsibility to own their infrastructure, software and processes - especially since this field is moving so fast and still so new (compared to 40+ of software development).

We have a great starting point if we don’t start at 0, but from where the DevOps mindset currently stands.

I couldn’t agree more. The space of ML and MLOps is still at a very early stage, and unfortunately it feels as if organizations have to re-learn all the lessons from software development that were learnt the last 20 years. Tools have a great chance here to bridge the gap a bit more harmonious than what we’ve seen in classical software engineering.

I'd second that, somewhat. The things I'd consider myself to be good at are the things I care about. I can still quite clearly distinguish between personal and "business" interests, and don't feel a blurriness between both if I do not want them to blurr, but it is the one thing that has helped me (apart from being at the right time at the right place) to progress in my career.

What I'd take away from what both godelski and me are saying (if I were you): Find topics that sound interesting to you, dig your heels in, learn about those, play with the concepts. 9 times out of 10 something will come out of it, and if its just yourself understanding better what you're actually looking for.

It really gives me great pleasure to find gems like this article on HN. I can't see a future in which I'd otherwise gained the insights I've now gained through reading (your?) piece about the complexity of blurring complex shapes. Thanks for that.

Indeed. Disregard my comment while I facepalm.