The Alt-Ac Job Beat Newsletter Post 23 (last for now)

Hi Everyone,

I started this newsletter to put to paper some of the advice I have been giving people one-on-one about the alt-academic job market for the past few years. I have written most of what I wanted to say -- you can use your skills you learned in your CJ PhD in various tech roles, applying broadly is a reasonable strategy if you don't have a good network to leverage, some basics of large language models tasks, and various little tech tips I found useful to better understand software engineering.

The last piece of advice is about the volatility in our current situation. Those who have traditional tenure track jobs may be looking at the federal government, or volatility in the private sector market, and think I want that stability of a tenure track job. I think that is mostly a false promise, as I have been involved with institutions in my career (and I am not that old) that both have reduced whole departments and frozen pay increases (from already low salaries). It is not in jest for me to say if your salary is double what it would be as an academic, you can work for one year, be laid off for a year, and still be in the same situation as you would be working at the traditional tenure track position.

The way to buffer yourself against the uncertainty is to increase your chances of landing roles -- level up your skills so that you can apply for technical roles in the private sector. Even with fewer openings in tech, in many urban markets there are still I am guessing around 1000 to 1 roles available in the private sector relative to traditional tenure track academic openings. And they almost uniformly pay better than the professor gigs.

Always feel free to send me a message with a question. And you never know I may start the newsletter back up if I have something more to say. I did try to start a recruiting service (which has not been successful, but that is OK). I do think there are still massive inefficiencies in the recruiting market, but will need to tackle that in the future.

JOBS

The job sheet for now is dormant. I encourage folks to check out the job sheet as an illustration for the types of jobs that are available that I think many criminal justice PhDs will be qualified for.

For an overview of job hunting advice I have given in the past:

CODE EXAMPLE

The last code example I want to share is using large language models for what is called structured output. For example you can submit a question to a large language model (ChatGPT, Anthropic, etc.) a question along the lines of:

I will give you a description of a burglary
please return json that determines the method
of entry, and the location of the entry
for example:

Offender pushed in the AC unit on the west side of the house.
return {'moe':'window', 'loc': 'side'}

The offender entered via the unlocked front door
return {'moe':'unforced', 'loc': 'front'}

The burglary narrative is ...insert narrative here...

And for Sonnet 3.5, if I submit in the insert narrative here part The offender broke in a window on the back door to enter the residence., it will return {"moe": "window", "loc": "back"}. A common business process I am coming across now is taking large amounts of unstructured text data (narratives, information in PDF files), and extracting that structured information the business wants.

The models have been improving over time, and are getting cheaper over time (I estimate if you did this process for 10,000 burglary narratives using Sonnet on AWS services, would be maybe around $30).

This is an example of using k-shot examples in the prompt as well. The context windows are so large for many of the new models (a 200k token context window means you can probably submit 100,000 words). You can basically just put a bunch of examples in the prompt.

Best, Andy Wheeler