iamnafets

No credential will be sufficient, this is basically an unsolvable enforcement problem. That doesn't obviate the utility of rules and norms, but there's no airtight system which will hold back AI generated content.

The warm container host is shared by multiple people. It basically runs docker with a small UI and a daemon which spins up new hosts, recycles them, and warms them when code changes happen.

Ideally when I open a new workspace, it opens in the context of that SSH session. So new splits open as new shells, claude's messages proxy through, etc etc.

Feel free to send me an email at stefan@hellointerview.com if you want to discuss. Love the problem you're trying to solve here.

Can’t wait for better SSH support. We have a warm container host that spawns a fresh container so we get nice sandboxing and isolation, but the UX has left us with basically the same challenges you solved here for one box.

One of the authors (Stefan) here. We created System Design in a Hurry after conducting hundreds of mock interviews and pinpointing areas where candidates tended to falter which could be fixed with some study. Happy to answer questions!

Should add liquidation preferences and allow sales below the last priced round. Seem to be more common these days!

Someone needs to add common LLM injections to the list.

Agreed. Scrum, in most cases, robs engineers of the agency they need to deliver maximum value. My teams did better work without it under the condition that the talent was motivated, technically proficient, and had the right incentives.

The obvious needle thread here is to not "make" engineers do anything, but hold them accountable to the results the business needs to see. The best teams are composed of empowered, accountable engineers who have the flexibility to do what they're paid to do.

Sounds like the problem is engineers aren't accountable for quality. Rather than prescribing a solution, these leaders should make sure incentives are correct in their organization.

Doesn't this line of argumentation undermine OpenAI's TOS which disallow training models on their output?

They are a YC company.

Still a major work in progress, feel free to send us any feedback stefan@hellointerview.com . Lots of work happening now.

This gave me PTSD from working in T&S. Impressive level of detail, nice work creating it!

Definitely. Much of it _isn't_. But the parent comment was referring to the waste of the process, which is necessarily the part that isn't directly useful in the job.

Similar in magnitude to the lost productivity of people optimizing for appearances rather than results at work. All prep is basically an arbitrage between the hard work of actually becoming better and the often easier work of preparation.

The ideal system would minimize this arbitrage, both for the sake of employers and candidates alike.

You can test this. Stratify users into groups with varying ad load and watch retention metrics. There's a bunch of 2nd-order effects that are hard to control, but you don't get to be a business doing billions in revenue without fine-grained understanding here.

Absolutely not. SAFE investments have liquidation preferences - the investors will get what's the left of the money back.

I think the problem is that the tradeoffs already exist. Most users would prefer more usable space or less money to a full history of their data.

You might be making the argument that the usability of immutable data is not there yet, but there are well-established theoretical costs of maintaining full history and I don't think they're within bounds of many real-world use-cases.

I generally feel the same way, but just to steel man the argument: would your manual code review process have caught this issue?

Sometimes we compare new things against their hypothetical ideal rather than the status quo. The latter is significantly more tractable.

Facebook Dangerous Content ML | Seattle | Full-Time | On-Site ($300k+ depending on experience)

I’m the hiring manager for Facebook’s Dangerous Content ML team and we have a unique opening in our group.

We build cutting-edge machine learning technologies used across the Facebook Family of Apps to keep our community safe and secure. Our work impacts the world and we’re focused on safety issues related to terrorism, hate organizations, human trafficking, and more. Along the way, we’re pioneering unique applications of ML from active learning and multi-modal live video classification to self-supervised learning and knowledge graph representations.

The team works alongside a cross-functional team of experts and partners who guide our development and provide subject matter expertise. You’ll have the opportunity to work in a ground breaking and fast-paced environment, taking on technical challenges at Facebook scale and devising technical strategies that have yet to be defined. The problems we’re solving are adversarial. The landscape is constantly shifting. Working at Facebook means taking on amazing technical challenges.

If you’re a senior ML engineer or scientist (4+ years) interested in applications that involve severe class imbalances, multi-modal data, and Facebook-scale requirements, I’d love to chat or buy you coffee. Email me smai@ (fb.com) with your resume and a brief introduction. Interview process would involve a technical phone screen followed by an onsite interview involving behavioral questions, coding, and both ML system design and research deep-dive.

For ML systems, it's an engineering mistake to deploy a complex model when you don't have a simpler baseline (e.g. does this outperform a basic n-gram model?). Similarly, it's a strategic mistake to deploy a deep learning model without assessing the baseline of human performance (including bias).

I see the problem of inexplicability as less salient than (1) responsible, informed deployments of models, and (2) ongoing measurement (especially against a human baseline).

You can deploy explainable models without (1) and (2) and end up with a much, much worse result.

This is mostly off-topic, but being able to identify or root out the value mismatches between people in a discussion (as you've done) seems like an essential ingredient to productive discussion.

I've noticed that people who can do this are disproportionately better at achieving their goals than those who can't, especially in engineering organizations. Exceptions of course for zealots and revolutionaries.

I've found Adele Cutler's presentation on random forests to be an outstanding resource for getting intuition of tree-based algorithms.

http://www.math.usu.edu/adele/RandomForests/UofU2013.pdf

Thinking about trees as a supervised recursive partitioning algorithm or a clustering algorithm is useful for problems that may not appear to be simple classification or regression problems.

I work in Forecasting for Amazon and I've often wondered the same thing. Almost any business with some (moderate) degree of uncertainty about the future could be "securitized". I think doing so in a way that preserves privacy, security, and is defensible against disintermediation could be valuable.

All that said, the company that outsources (and that's really what you're proposing) such a core component of their business is probably taking on way too much risk.

Staffing up a team from 4 to 12. Same role, but different levels and responsibilities.

Amazon New Product Demand Forecasting | Seattle | Full-Time | On-Site ($130-$250+ depending on experience)

Amazon's New Product Demand Forecasting team is responsible for one of the most challenging problems in supply chain optimization: predicting sales for products that have no sales history. This is a uniquely creative space in Forecasting requiring our machine learning models to capture both the nuances of the global consumer marketplace as well as customer behavior on Amazon.

Our team works closely with research scientists to invent new ways to make use of novel data, solve hard engineering problems around scaling and performance in predicting for tens of millions of products, and iterate quickly in order to stay on the cutting edge.

I'm looking for an experienced software developer (sorry, no university hires at this time!) that is comfortable with big data and machine learning and can:

* Design systems that provide a stable base for innovation in a rapidly changing business

* Improve Forecasting algorithms through data-driven analysis and experimentation in our Scala/Spark environment

* Optimize for scalability and performance of both distributed computations and near-metal C++ code

* Learn quickly and keep up with a rapidly changing machine learning and big data landscape

* Communicate their ideas clearly with all members of a diverse team

If this sounds interesting, as the hiring manager I'd love to chat or buy you coffee. Email me (Stefan) at smai@ (amazon.com) with your resume and a brief introduction. (Interview process is 1 phone screen and onsite interview with whiteboard coding and behavioral questions about your experience.)

P.S. Big shout out and thanks to the HN community for the "Who is hiring?" threads! I've gotten to meet or email with nearly 100 very talented engineers and scientists (and hired several) over the last 2 years. I'm now down to 1 more opening on my team. I'll say this: if you think you're not qualified -- send me an email anyways. I've been flabbergasted with the number of strong engineers who preface their email with "I'm probably not a fit". Imposter syndrome is real -- send me an email anyway and let's at least have a chat!

At this time, yes. But if you're at a place where relocation is an option, Seattle is an awesome place to live!

Amazon New Product Demand Forecasting | Seattle | Full-Time | On-Site ($130-$250+ depending on experience)

Amazon's New Product Demand Forecasting team is responsible for one of the most challenging problems in supply chain optimization: predicting sales for products that have no sales history. This is a uniquely creative space in Forecasting requiring our machine learning models to capture both the nuances of the global consumer marketplace as well as customer behavior on Amazon.

Our team works closely with research scientists to invent new ways to make use of novel data, solve hard engineering problems around scaling and performance in predicting for tens of millions of products, and iterate quickly in order to stay on the cutting edge.

I'm looking for an experienced software developer that is comfortable with big data and machine learning and can:

* Design systems that provide a stable base for innovation in a rapidly changing business

* Improve Forecasting algorithms through data-driven analysis and experimentation in our Scala/Spark environment

* Optimize for scalability and performance of both distributed computations and near-metal C++ code

* Learn quickly and keep up with a rapidly changing machine learning and big data landscape

* Communicate their ideas clearly with all members of a diverse team

If this sounds interesting, as the hiring manager I'd love to chat or buy you coffee. Email me (Stefan) at smai@ (amazon.com) with your resume and a brief introduction. (Interview process is 1 phone screen and onsite interview with whiteboard coding and behavioral questions about your experience.)

Amazon New Product Demand Forecasting | Seattle | Full-Time | On-Site ($130-$250+ depending on experience)

Amazon's New Product Demand Forecasting team is responsible for one of the most challenging problems in supply chain optimization: predicting sales for products that have no sales history. This is a uniquely creative space in Forecasting requiring our machine learning models to capture both the nuances of the global consumer marketplace as well as customer behavior on Amazon.

Our team works closely with research scientists to invent new ways to make use of novel data, solve hard engineering problems around scaling and performance in predicting for tens of millions of products, and iterate quickly in order to stay on the cutting edge. I'm looking for an experienced software developer that is comfortable with big data and machine learning and can:

* Design systems that provide a stable base for innovation in a rapidly changing business

* Improve Forecasting algorithms through data-driven analysis and experimentation in our Scala/Spark environment

* Optimize for scalability and performance of both distributed computations and near-metal C++ code

* Learn quickly and keep up with a rapidly changing machine learning and big data landscape

* Communicate their ideas clearly with all members of a diverse team

If this sounds interesting, as the hiring manager I'd love to chat or buy you coffee. Email me (Stefan) at smai@ (amazon.com) with your resume and a brief introduction. (Interview process is 1 phone screen and onsite interview with whiteboard coding and behavioral questions about your experience.)