How to Have a Bad Career – David Patterson (2016) [video]

I’ve implemented LLM based “intent processing” instead of old school “give it a list of every possible phrase that someone could use to activate an intent” for call centers for a couple of states and a couple colleges - I worked for AWS ProServe directly in the pub sec division and now work for a third party consulting company.

Think of an intent as something like asking Siri for driving directions, setting an alarm etc.

One of my specialties is Amazon Connect for hosted call centers - based on the same service that Amazon uses internally. Think of the difference between how Google Assistant and Alexa worked pre LLM and how Siri works today compared to tool calling that modern LLMs used.

That’s just using an LLM in a product.

As far as how I use it everyday? I am a staff consultant and in the previous times, I needed at least a couple of developers under me just to get the grunt work done on time. Now I can just treat Claude Code and Codex as faster, more accurate ticker takers.

Of course I’m not going to dox myself (more than I probably already have) or break any NDAs

And before the gate keeping starts about “I must be inexperienced”. My first time coding was on a 1Mhz Apple //e in assembly and BASIC in 1986. During the first decade and a half of my professional career, I programmed in C targeting everything from mainframes to PCs to Windows CE ruggedized devices.

I think it depends on the interests of the researcher. If a researcher is comfortable being a “brain for hire” who is comfortable solving research problems that are driven by business needs and where there needs to be short-term or medium-term results, then I think there are plenty of opportunities at large companies, including the FAANGs. I find research more fun than software engineering, but researchers are far from immune from pressures to ship.

If a researcher is more interested in curiosity-driven work and who wants to work on a longer time frame, I’m afraid that there’s no place in industry, except for maybe Microsoft Research (which I’ve heard changed under Satya Nadella), that supports such work. The days of Bob Taylor-era Xerox PARC and Unix-era Bell Labs ended many decades ago, and while there were still curiosity-driven labs in industry well into the 2010s, I have witnessed the remainder of these old-style labs change their missions to become much more focused on immediate and near-immediate business needs.