hectormalot

Stellar | Software Engineer | Amsterdam, Netherlands, EU | Onsite/Hybrid | Full-time

We're building AI voice agents for enterprise contact centers. We focus on remarkably human-like conversations, regulated environments and support for non-English markets (Dutch, German, French, etc.).

All our founders can code, but we're growing and getting busy quickly. So, we are looking for a (senior) Software Engineer: Go/TypeScript (Encore.dev/React). If you're strong in other languages, entrepreneurial and eager to learn, we'd still love to talk. Build core platform features, guardrails, integrations, and tools that help our AI agents handle millions of customer calls.

You'll thrive here if you:

- Want high ownership and direct customer impact

- Prefer building directly with users vs. in isolation

- Enjoy technical challenges at the intersection of AI/audio/enterprise

- Have the ambition to grow into a technical leadership role as the team grows

More information: https://www.stellarcs.ai/careers/software-engineer

EU work authorization is required. No visa sponsorship available.

Apply: e-mail is in my profile, please indicate HN in your e-mail.

Stellar | Implementation Lead, Software Engineer | Amsterdam, Netherlands, EU | Onsite/Hybrid | Full-time

We're building AI voice agents for enterprise contact centers. We focus on remarkably human-like conversations, regulated environments and support for non-English markets (Dutch, German, French, etc.).

1. Implementation Lead: Lead customer deployments from POC to production. Technical understanding required, coding skills not a requirement. You'll work directly with enterprise clients to integrate our AI agents into their existing contact center infrastructure.

2. Software Engineer: Go/TypeScript (Encore.dev/React). If you're strong in other languages, entrepreneurial and eager to learn, we'd still love to talk. Build core platform features, guardrails, integrations, and tools that help our AI agents handle millions of customer calls.

You'll thrive here if you:

- Want high ownership and direct customer impact

- Prefer building directly with users vs. in isolation

- Enjoy technical challenges at the intersection of AI/audio/enterprise

EU work authorization required. No visa sponsorship available.

Apply: e-mail is in my profile, please indicate HN in your e-mail.

One reason I could think of is that they may return the database (or cache, or something else) response after generating and storing the OTP. Quick POCs/MVPs often use their storage models for API responses to save time, and then it is an easy oversight...

Having some experience with both, I think they are quite different. N8n looks quite polished and seems primarily concerned about connecting pre-made blocks. There are custom code blocks (JS and Python only, with limited ability to import libraries), but it’s not something you’d use by default. I thinks it great for less-technical users when compared to windmill.

Windmill OTOH supports a bunch of programming languages for steps (Go, Rust, Python, TS, etc.) and seems to have a much more “code first” approach. Reusable blocks are more like code templates compared to n8n.

Hard to say which is better. I really like the ability in windmill to just write code for each step and it comes across more powerful, but it feels less polished and intuitive when compared to n8n.

The story is from a co-host with Boris Johnson for some award ceremony. It’s a great read: https://m.facebook.com/story.php?story_fbid=2449074521979085...

With Johnson I at least had the impression that he understood the showmanship aspect of it really well. Less so with Trump, at least it seems less polished.

I think that is a result of the economics with tips.

Restaurant owners need to balance between overstaffing (and thus spending too much) and understaffing (and having service suffer). The optimal point is different if your opportunity cost is $2 vs $15 per hour.

Thus, in a US restaurant, on average, you’ll see more waiters than in most EU places, given the same amount of tables.

I think people also heavily underestimate what 1000s of years means. This type of storage has to survive 3x as long as the Egyptian pyramids. The problem is not just technological. At those timespans you can’t assume the country you live in - or the language you speak - to still exist.

In some dimensions I recognize the slow down in how fast new capabilities develop, but the speed still feels very high:

Image generation suddenly went from gimmick to useful now that prompt adherence is so much better (eagerly waiting for that to be in the API)

Coding performance continues to improve noticeably (for me). Claude 3.7 felt like a big step from 4o/3.5. Gemini 2.5 in a similar way.compared to just 6 months ago I can give bigger and more complex pieces of work to it and get relatively good output back. (Net acceleration)

Audio-2-audio seems like it will be a big step as well. I think this has much more potential than the STT-LLM-TTS architecture commonly used today (latency, quality)

Oh my bad, I thought you referenced contractors in general. In that case I agree: agreements are to be honored.

I don’t think that’s shady? When I was hiring contractors it was always project based with mutual understanding things could end quickly if the project or collaboration didn’t work out.

Yes, they get paid 1.5-2x, and that also prices in that it’s not always 100% utilization. Only once had a contractor oppose that, but that was in the context of (severe) underperformance.

Yes, by API usage. I think you can’t limit which files it sends.

Likely because it is blended with cached token pricing, which is at $0.30/million. You can use ‘group by’ in the usage portal to see the breakdown.

I’ve read a few of these in depth crash analyses: https://admiralcloudberg.medium.com/

I’ve realized that fires are a way bigger issue than you might imagine after a crash — things can go south really quickly. Multiple stories of planes going from “fire outside” to “people suffocating and burning to death in their seats” in minutes. Here’s one of a 737 in Manchester, taxied of the runway intact, 55 people died: https://admiralcloudberg.medium.com/fire-on-the-runway-the-m...

I might have taken my laptop bag in such a case out of habit before reading these stories, not so much now.

It’s a bit more complicated than that.

At the point I make a loan, 2 things happen on my balance sheet: I have a new liability to you (the increased balance in your account), and I have a new asset (the loan that you’re expected to pay back). They cancel each other out and it therefore seems as if I’m creating money out of thin air.

However, the moment you actually use that money (eg to buy something), the money leaves the bank (unless the other account is also at this bank, but let’s keep it simple). Liabilities on the balance sheet shrink, so assets need to follow. That needs to come from reserves because the loan asset keeps its original value.

The reserve comes from the bank, not from you. Added layer here: Banks can borrow money from each other or central banks if their cash reserves runs low.

Finally: it tends to be the case that the limit on lending is not the reserves, but on the capital constraints. Banks need to retain capital for each loan they make. This is weighed against the risk of these loans. For example: you could lend a lot more in mortgages than in business loans without collateral. Ask your favorite LLM to explain RWAs and Basel III for more.

ABN AMRO | Amsterdam, Netherlands | AI Developer | HYBRID | Full Time | Python

We're looking for experienced people who want to help with our GenAI program. Roles range from project management to engineering to data roles, although there is less hard data science where LLMs are involved.

We’re part of Strategy & Innovation and we work on new products and new technology roll-outs across the bank. The GenAI roll-out is one of our primary responsibilities.

We’re looking for people with a few years of work experience. This may also be in a different field, entrepreneurship or for example consulting. If you’d be comfortable taking 1-2 juniors under your wing you’re in the right ballpark.

Please note that we're not sponsoring visa's for applicants outside of the EU.

Mention HN when reaching out. My email is in my profile.

Depends on your perspective. I think it’s like front-end development likely means you work _with_ React, not that you’re writing front end frameworks from scratch. In a similar way AI devs in our team work with LLMs, but they don’t create them from scratch.

Totally possible. About 30-40% of the people in our AI team don’t have a formal AI background. Especially with LLMs a lot of work has shifted towards “data literate software engineering”. We call them AI engineers / AI developers. Good development skills are very transferable to those roles.

Feel free to reach out if you’re in the EU (email in profile), we’re hiring. Also happy to give some pointers on how to approach these conversations.

…yet.

Negative energy prices are a relatively new phenomenon, and it attracts attention because of its counterintuitive nature (I can get paid to use electricity?!?). Yet, it doesn’t happen that often. If I look at the Dutch market, this mostly happens on sunny+windy Sundays.

I think 2 things are happening that will make arbitrage feasible:

1. As solar production capacity increases, so will opportunities for arbitrage

2. As battery costs keep going down (capacity learning curve), so will the investment to do arbitrage.

At some point the curves for (1) and (2) will cross convincingly enough that investing in storage is just good business, and people will invest. If it’s anything like solar (home batteries), adoption might actually go very quickly because decision making would be decentralized, small investments, and without permits.

We’re automatically logging and summarizing calls. I’m in a regulated industry so it’s not an optional activity and there are quite some requirements to how and what we log.

Interesting thing we learned is that agents tend to log during the calls, not afterwards. We now see (qualitative feedback) that they are less busy logging and therefore have more attention for the client, and (quantitatively) we see the calls are getting shorter and people with the tool are doing more calls per day. We do many millions of calls a year, so it sums out to a good number.

Similarly, we have processes with 100s analysts with very high standards to their outputs. The traditional way is to have QA teams review and provide feedback for a few rounds. We’re introducing AI for the first round(s) of feedback to shorten the cycle time reduces context switches) and have the QA teams focus on final reviews.

But I get your point. RAG knowledge bases for experts are in a hard spot. After a few months of employment, the experts tend to know the general knowledge well. As a result, the RAG-bot mostly gets questions about exceptions and niches, where it doesn’t perform very well, and mistakes might be expensive.

Interesting way to look at it! At least looks like companies are investing with some expected future return that is not yet there today.

Other thoughts:

1. Current revenues might be a bit higher than you calculated. E.g. I’m not sure if copilot and azure OpenAI service are fully included in those revenues, and those might be relevant figures at this scale.

2. The 10-100x growth might actually materialize. Corporates are much slower to adopt and scale a technology than people might expect. As a result, many big potential users are only at the very beginning of adopting AI. (I am assuming there are valuable applications for them to use AI for)

I once spoke with some engineers at Monzo and at the time (2018) financial transactions were handled via a combination of Kafka and Cassandra.

ABN AMRO | Amsterdam, Netherlands | AI Developer | HYBRID | Full Time | Python

We're looking for experienced people who want to help with our GenAI program. Roles range from project management to engineering to data roles, although there is less hard data science where LLMs are involved.

We’re part of Strategy & Innovation and we work on new products and new technology roll-outs across the bank. The GenAI roll-out is one of our primary responsibilities.

We’re looking for people with a few years of work experience. This may also be in a different field, entrepreneurship or for example consulting. If you’d be comfortable taking 1-2 juniors under your wing you’re in the right ballpark.

Please note that we're not sponsoring visa's for applicants outside of the EU.

Mention HN when reaching out. My email is in my profile.

I wouldn't underestimate how much a modern machine with a bunch of RAM and SSDs can do vs HDFS. This post[1] is now 10 years old and has find + awk running an analysis in 12 seconds (at speed roughly equal to his hard drive) vs Hadoop taking 26 minutes. I've had similar experiences with much bigger datasets at work (think years of per-second manufacturing data across 10ks of sensors).

I get that that post is only on 3.5GB, but, consumer SSDs are now much faster at 7.5GB/s vs 270MB/s HDD back when the article was written. Even with only mildly optimised solutions, people are churning through the 1 billion rows (±12GB) challenge in seconds as well. And, if you have the data in memory (not impossible) your bottlenecks won't even be reading speed.

[1]: https://adamdrake.com/command-line-tools-can-be-235x-faster-...

My sense is that finetuning might still have a role if your use is high-volume and with a narrow/specific goal. For example, we have GPT based summaries of our customer contact calls. It's fairly high volume (a typical bank will handle millions of calls a year).

We're considering finetuning for 2 reasons:

(1) the current system prompt with instructions is getting quite large* and we see more challenges for GPT to stick to the instructions. We could finetune with our historic summaries and a simplified prompt to slightly improve the performance of the summaries / lower token count (performance). The idea would be to then continue improving our system prompt again from that new starting point. Exploration still to be done though.

(2) we might be able to finetune a much smaller model to do what the larger model is currently doing (cost, sustainability)

* The larger instruction prompt is because there are lots of specific needs for the summaries (e.g. how to write the summary, what to exclude (health, etc.), what to include (actions taken), and we e.g. give examples of good vs bad summaries to improve performance.

Depending on your experience the roles fall into levels 10, 11 and 12 of our CLA[1]. That translates to €64k up to €122k annually if you work 40h/wk.

[1]: https://www.fnv.nl/getmedia/88e7fefb-7035-40bb-83bd-9968906f...

Note: If you want a DM, you should probably put your contact details in your profile ;)

ABN AMRO | Amsterdam, Netherlands | AI Developer | HYBRID | Full Time | Python

We're looking for experienced people who want to help with our GenAI program. Roles range from project management to engineering to data heavy work, although there is less hard data science where LLMs are involved. We’re part of Strategy & Innovation and we work on new products and new technology roll-outs across the bank. GenAI roll-out is one of our primary responsibilities.

We’re looking for people with a few years of work experience. This may also be in a different field, entrepreneurship or for example consulting. If you’d be comfortable taking 1-2 juniors under your wing you’re in the right ballpark.

Please note that we're not sponsoring visa's for applicants outside of the EU.

Mention HN when reaching out. My email is in my profile.

I largely agree with your sentiment. Go’s simplicity is what makes it such a useful tool for me. It’s worth protecting, and that means setting a very high bar for proposals that add new things to the language.

However, 2 things I would be enthusiastic about if it got included in the language: - having ‘?’ As syntactic sugar for ‘if(err != nil) …’. Would make code more easily readable, and I think that is a benefit for programmers trying to keep things simple. - Sum Types. I’ve had a few cases where this would’ve been very useful. I consider the ‘var state customtype = iota’ a bit too easy to make mistakes with (eg exhaustive checking of options).

Like generics, I hope that when that happens, they take a very deliberate approach on doing it.

Not exactly the same feature set, but I’ve replaced Dropbox with Syncthing (open source) and it has been very reliable for my intended use (sync files across my devices).

I also run it on my home server for a central sync point. Don’t use it much on mobile, but works well enough if I do need a file. Not a good solution for sharing with others though.

ABN AMRO | Amsterdam, Netherlands | Full Stack Developer | HYBRID | Full Time | Java and Javascript

We're looking for full stack developers with experience in Javascript (VueJS) and Java (Spring Boot). The Innovation Engineering department works on new products and supports new technology roll-out across the bank. GenAI roll-out is also one of our responsibilities (for which we require Python).

Please note that we're not sponsoring visa's for applicants outside of the EU.

Vacancy here: https://www.werkenbijabnamro.nl/vacature/6190/full-stack-eng...

Feel free to reach out to the e-mail in my profile for more information

If you want to use the CLI, Visidata might be what you want. It does have a bit of a learning curve. Beyond that, I've found it quite handy to do quick data explorations. e.g. there are shortcuts for histograms, filtering, x-y plots, etc.

[1]: https://www.visidata.org