6

Last year I took a job doing machine learning for a small company right after finishing my masters in which I was doing mostly AI research and managed to publish some stuff. At work, I'm the only one working with AI meaning that I'm my own lead and have complete freedom of how-and-what to do. Given that during my studies I had very good and qualified professor as my supervisor, whenever I would come into a crossroads with my research or simply was having a hard time grasping some new concepts she would help me out but not without me putting the effort of trying my best first. Since I got my new job, I'm feeling very doubtful of my findings, approaches, experiments, etc, etc... I notice myself wishing to have some sort of guidance or at least another person in my team bounce ideas off.

I’m thinking of telling my manager about it, and ask him to bring another person into the team. I’m not so afraid of losing my job but to seem unfit or not capable enough for the position.

Has anyone been in a similar position of having to asking management to hire people to help you out? What was their reaction? Did it they treated you any differently?

5
  • 5
    If you are still in touch with your prof, ask her for general (not problem-specific) advice. However, be prepared that one reason you have been hired is independence. Careful what you wish for: if they can hire someone for you to learn from, they can as well just keep them and fire you. "Bounce ideas off" - if they have to get 2 people to get the work from you (plus perhaps the new person), they can as well aim for 1 independent person or 2 of them. So, careful how you present your request. May 12 '20 at 16:42
  • 4
    Consider mentorship outside of your workplace. Are you a member of any professional organizations that may be able to help?
    – Myles
    May 12 '20 at 17:42
  • @CaptainEmacs Yeah, I'm aware of that. But as I mentioned, if they decide to let me go because of that I'm at peace with it. At this point I find more overwhelming the feeling of uncertainty about my knowledge that loosing a gig... I mean I could go back to software engineering if I'm not cut for research I guess. May 12 '20 at 17:48
  • 1
    @Myles That's a great advice actually. I'll try to find a Data Science/Machine Learning community in my city. Thanks! May 12 '20 at 17:49
  • If you are interested in AI research send me an email to basile@starynkevitch.net mentioning the URL of your question May 12 '20 at 18:18
4

You should absolutely discuss this with your manager. Your manager should have a very good understanding of what you're experiencing right now. That being said, be prepared to be told "We can't do that. It's not in the budget."

These things are often driven by product roadmap and budget. There are a lot of considerations that go into hiring/acquiring more headcount, and if it doesn't apply to the company's roadmap or strategic initiatives it simply won't be freed up.

Your best option at this point is to take account of what the company's goals are and how the current allocation is not meeting those needs. You need to come up with the specific costs and and values associated with a single headcount. What will this person provide to the company. While I understand you want some value for your initiatives and growth, the company can't redirect resources without getting some very well defined value for the company moving forward. If they're going to invest $X in something, there needs to be $X times 4 value provided in product or capabilities for customers.

It's unlikely you'll be treated differently really. Your manager will simply be aware of your concerns, and they might modify some of their action items to make sure they're addressing how you feel and progress. They certainly won't look at you negatively unless you just whine and complain about things not being what you want. If you show up with data, propositions, recommendations and goals they'll respect you much more for that.

1
  • I'll talk to them later this week, it would give me time to gather all the material to get my point across. Being realistic I don't think we are in a good enough position to bring someone else in due to the covid-19. May 12 '20 at 17:59
4

You should read the Mythical Man Month : Essays on Software Engineering book and the Bullshit Jobs: A Theory book. Of course, be aware also of Peter principle (it is a fixed point analysis).

Both are very relevant to your situation.

Brook's book (mythical man-month) demonstrate that communication between human software developers and their management is what makes most software development projects fail or be late (and in many cases, adding more manpower makes the project more late or worse). That was true in 1970s with punched cards computer and software, that is still true in 2020s with multicore computers with webcams. IIRC about half of today software projects are somehow "failing" - either buggy or not implementing correctly all the initially desired features.

Graeber's book (Bullshit jobs) argues the existence and societal harm of meaningless jobs. He contends that over half of societal work is pointless (and you can include a lot of software development in that classification), which becomes psychologically destructive when paired with a work ethic that associates work with self-worth.

Of course you need (if the success of your software development job matters to you) to calmly suggest to your manager that additional labour might help. He might have constraints you are not aware of (e.g. budget, timescale, client, etc...) and you should understand them.

The level of trust between you and your manager is specific to your particular situation: some managers are good persons, other are not. Some are proud to make their team happy and productive, others are just dreaming to climb the company organizational ladder. Most are in between.

You could (or not) consider writing - or contribute writing -, at home on your own computer, some open source software (e.g. on github or gitlab), perhaps RefPerSys. In some workplaces and legal systems this could even be part of your job (e.g. most of GCC or Linux kernel contributors are paid for their work). This paper on simple economics of open source is explaining why that could be your interest (such as your personal reputation in professional or IT related circles, or learning new IT technologies for a future job transition, and at the same time have fun).

If your professor was nice with you and you can and want to keep in touch with him, you might suggest your boss to pay some consultancy services to him.

Don't forget that the goal of a company is to maximize profit. Human staff (notably software developers) is always an implementation detail. And most of us software developers have some kind of impostor's syndrome (me included, and I am grandfather of 7 with a PhD in AI...). Remember of course Rice's theorem related to the Halting Problem. Read Gödel, Escher, Bach if you didn't read it... See also this draft report and the references inside it.

Alternatively (and related to the ongoing Covid 19 pandemics) ask politely for additional delays in your work.

1
  • You're right. Good thing is that my manager is actually very nice and we've both acknowledged that, for the most part, I'd be learning more as I go on with my work there (such is the way of research). Their expectation is quite realistic, I guess I panic at not having some form of validation other than performance metrics of my models. Anyways, I'll let him know how I feel and maybe we can revisit expectations and such. PD: I'll definitely take a lookt at those books! May 12 '20 at 17:55
2

Presumably, the company is interested in AI-related skills enough to hire you. Presumably, you know about AI more than anyone else at the company. At the same time, you need another person to bounce your ideas from.

Run a regular AI-related journal club in your company

Pros for you: you get to read relevant articles, you get more people who understand how AI works, you get people to talk with about AI.

Pros for the company: the bus factor in the AI direction will be higher than the current value of one.

Depending on specific topics you discuss and confidentiality levels, you might want to invite your past collaborators from academia or members of your local AI community. Figure out with your manager if the journal club should be weekly, biweekly or whatever.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .