HN Explorer

Ask HN: what problems in technical and scientific computing are worth solving?

10 points by mynegation ↗ HN
TL;DR - if you work with high-performance computing what problems would you like to be solved in that area?

I work in finance and while there are things that demand a lot of expertise in high-performance computing, low-latency computing and big data and are technologically challenging and very interesting, I sometimes feel that there could be a better application for my skills.

My question is to HNers who work in fields that need a lot of computing horsepower: bioinfo, pharma, aerospace, geological survey, high energy physics, actually any kind of industry that does a lot of analytics.

What, in your opinion, are challenges in the area of high-performance computing, that - if solved - would make your life much easier: be it algorithms, infrastructure, tools, or workflows?

5 comments

[ 3.1 ms ] story [ 22.0 ms ] thread
paulsutter
Generally speaking, high performance computing challenges aren't valuable. And I say this as a guy whose last startup makes 50 million decisions a second and processes 10 petabytes a day (Quantcast).

Making the technology scale was the -fun- part. The company gets paid for solving a specific business problem. A business problem that costs the world $600B a year (advertising).

It's hard to start with a technical ability and find a startup opportunity. That is how we decided to start Quantcast - we had a spreadsheet of business ideas that might leverage our technical backgrounds. But it's a difficult path.

It's easier to start with basic consumer or business needs and design a solution generally. The best business ideas you come up with are not likely to require the cool/fun high performance skills you (we) have.

Murphy's law or something.

strlen
> It's hard to start with a technical ability and find a startup opportunity. That is how we decided to start Quantcast - we had a spreadsheet of business ideas that might leverage our technical backgrounds. But it's a difficult path.

Really interesting point. I am much in the same boat as you, I am interest in working on highly technical, mostly systems related challenges.

However, I know that the goal of a company is to generate return for their shareholder. At the mean time, startup is a lot of hard, thankless, and grungy work (that makes tooling around with autoconf seem fun by comparison): unless you're genuinely passionate about the problem(s) you're solving, it isn't worth it.

So until a deeply technical idea strikes me, I'll hold off on taking that jump.