Ben Koller

Podcast: MLEngineered

Tue, 02 Mar 2021 00:00:00 +0000

You train it, you run it

Fri, 05 Feb 2021 00:00:00 +0000

The quote “You build it, you run it” is among the most influential quotes in modern software development. It has guided the conception of high-performing software teams since its inception. Pick any successful, software-driven business of today, their mentalities will all be at least influenced by the sentiment.

It’s origins go all the way back to 2006, when the CTO of Amazon, Werner Vogels, gave a seminal interview:

“Giving developers operational responsibilities has greatly enhanced the quality of the services, both from a customer and a technology point of view. The traditional model is that you take your software to the wall that separates development and operations and throw it over and then forget about it. Not at Amazon. You build it, you run it. This brings developers into contact with the day-to-day operation of their software. It also brings them into day-to-day contact with the customer. This customer feedback loop is essential for improving the quality of the service.”

A most unusual year

Sat, 26 Dec 2020 00:00:00 +0000

As this is the end of the year, it’s a great chance to remind yourself: how did we get here?

Let me beginn with a flashback to 2019. As a company, we’re focussed on optimising remaining useful live of industrial assets through clever use of Machine Learning for predictive analysis, root-cause analysis and other forms of reasoning. We managed to secure a few big projects and very promising POCs, and across the board we were able to show good results. One of our projects even got government funding, providing a nice runway going forward.

TMLS Summit 2020: The tale of a thousand pipelines

Thu, 19 Nov 2020 00:00:00 +0000

This is an Ops-guy’s story that tackles the process of building scalable deep learning pipelines for hundreds of model training on giant time-series datasets and on how it helped saved 80% of the cost along the way.

MLOps: Learning from history

Mon, 09 Nov 2020 00:00:00 +0000

TL;DR: Running successful ML teams and projects requires cross-functional collaboration within the executing team.

That sentence alone does not help anyone. It feels unsubstantiated, and someone playing bullshit bingo might luck out just on that one sentence - and yet, it’s true. Why? Because it’s taking the main driver behind the DevOps revolution in software development and applies it to Machine Learning.

If you’re familiar with what DevOps stands for, where it came from and why it’s such a game-changer, feel free to skip to the conclusion.

12 factors of reproducible Machine Learning in production

Mon, 28 Sep 2020 00:00:00 +0000

The last two decades have yielded us some great understandings about Software Development. A big part of that is due to the emergence of DevOps and it’s wide adoption throughout the industry.

Leading software companies follow identical patterns: Fast iterations in software development followed by Continuous Integration, Continuous Delivery, Continuous Deployment. Every artefact is tested on its ability to provide value, always has a state of readiness and is deployed through automation.

A case for declarative configurations for ML training

Sun, 17 May 2020 00:00:00 +0000

No way around it: I am what you call an “Ops guy”. In my career I admin’ed more servers than I’ve written code. Over twelve years in the industry have left their permanent mark on me. For the last two of those I’m exposed to a new beast - Machine Learning. My hustle is bringing Ops-Knowledge to ML. These are my thoughts on that.

Deploying software into production

Hundreds of thousands of companies deploy software into production every day. Every deployment mechanism has someone who built it. Whoever it was (The Ops Guy™, SRE-Teams, “Devops Engineers”), all follow tried-and-true paradigms. After all, the goal is to ship code often, in repeatable and reliable ways. Let me give you a quick primer on two of those.

Immutable Infrastructure

Thu, 30 Mar 2017 00:00:00 +0000

Immutable Infrastructure describes the separation of infrastructural state and application. Consecutive deployments spin up unique and explicitly defined infrastructure instead of replacing application artifacts on already existing infrastructure.

In the desire to deliver working software with each deploy without downtimes lie challenges immutable infrastructure attempts to solve.

Background

Startups in their early stages tend to neglect infrastructural design in favour of perceived speed in development. A prototypical production environment is focused on a single server. Backups are added as an afterthought and revolve around cronjobs. Deployments are a concatenation of manual steps and increase in risk each time. Downtime is almost always necessary during deployments. Even distributed designs suffer from the risks introduced by consecutive changes to their infrastructure.

Monitoring

Wed, 15 Feb 2017 00:00:00 +0000

Already a while back I got introduced to my current employer: Talentry. They, and thus me, try to make employee recommendation the #1 recruiting source. It’s a mission easy to align with as I strongly believe in it. Luckily, so do other companies, as the customer base proves. In the past months growth fell upon Talentry, and the platform’s infrastructure became due to an overhaul. This is where I step in. Along the way I’ll shed some light upon projects and organisational developments. Parts might be left out, parts will be intentionally vague, nonetheless the lessons learned will prevail.

Growth advice: All advice is about context

Mon, 12 Dec 2016 00:00:00 +0000

I worked in multiple companies in my live. I saw reasonable high evaluations of eight digits and more. I’ve had ample opportunity to examine organisational voodoo, and surely you have seen incomprehensible things happen in companies, too. Great design decisions, good management styles, innovative IT architecture, and all their negative counterparts. Growth advice will be my section to talk about the trapdoors of growing a business, and how to avoid them.

Interconnect VPCs with VyOS

Sun, 01 Nov 2015 00:00:00 +0000

TL;DR: Stylight, a company I used to work with, runs a number of VPCs in multiple regions, and only partial interconnectivity existed. After adding a set of new VPCs in 2015 for services that require a stable and performant interconnection a new solution was required. We used VyOS and the AWS VPNs to pull it off.

I’d like to tell you a joke about our VPN setup but it’s virtually private.

About

Mon, 01 Jan 0001 00:00:00 +0000

This is a blog, like everybody else’s blogs. However, it’s my blog, which makes it at least somewhat different than everybody else’s.

Things written about usually take their origins in my professional career. I started as a SysAdmin, picked up a casual interest in programming and settled on helping startups grow their infrastructure to accommodate the success they deserve. And because of an insatiable hunger and great people around me, I tend to “entrepreneur things”.

Imprint

Mon, 01 Jan 0001 00:00:00 +0000

Imprint

Sole responsible dude: me (Benedikt Koller).

Links may contain all kinds of shit, but because they’re hosted elsewhere you can’t blame me. Content here is mine, and if it isn’t I’ll make sure it’s abundantly clear.

Contact: sayhi@benkoller.de

I respect your privacy and don’t track anything. No external javascript, actually not even internal javascript. No hidden bullshit. Only you, your browser, your ISP and anyone with a BGP hop between you and this server will know you were here — I won’t track you.

Ben Koller

Podcast: MLEngineered

You train it, you run it

A most unusual year

TMLS Summit 2020: The tale of a thousand pipelines

MLOps: Learning from history

12 factors of reproducible Machine Learning in production

A case for declarative configurations for ML training

Deploying software into production

Immutable Infrastructure

Background

Monitoring

Growth advice: All advice is about context

Interconnect VPCs with VyOS

About

Imprint

Imprint

GDPR / Data privacy