It’s hard to over-emphasize the information asymmetry in job offers in AI. On one end, there are PhD grads straight out of school, with a tendency to self-reject / self-select based on what they perceive to be fair, and with little to no training on navigating the non-technical aspects of the recruitment process (understandably so!). And on the other, there are recruiters, who have significantly more data and are experts at their job, which gives them unmatched leverage in offer negotiations. Compound that with the lack of incentives for companies to publicly release pay data, the status quo is unlikely to change by itself. Some people are better informed than others based on how connected they are to the right set of people. This has significant consequences on various outcomes of the job search (e.g., role, level, compensation). Based on data from AI Paygrades, overall compensation can be negotiated up to as much as 2.5x the first offer. Some candidates know this, others don’t. Those who don’t, won’t know to negotiate.
That’s where services like aipaygrad.es and Ralph come in. Transparency is key to dismantling information asymmetry. aipaygrad.es aims to decrease information asymmetry by publicly releasing anonymized job offer data, similar to levels.fyi but for AI in particular. The intent is for candidates to use this data as reference points when comparing job offers. Ralph takes this one step further, empowering candidates with negotiation expertise so they can navigate the myriad of ways companies negotiate. Think real estate agents but for job offers.
The point here isn’t just to maximize pay, it’s to level the playing field, for both sides to have data and experience to make informed decisions, and not let biases manifest in pay disparities.
I’ve personally benefited from and enjoyed my interactions with Brian (from Ralph) during my job hunt. Brian brought significantly more experience to the table; had great suggestions on what questions to ask, and how to navigate the whole process. Going in, I expected the perception of research quality / productivity to be the primary factor in the compensation a company offers. Having gone through the process, that doesn’t seem to play as big of a role. Of course, you need to clear the interview process to get the offer in the first place, and offers from multiple companies can give you additional leverage in negotiation. But once past that stage, everyone in the process just seems to be doing their job. You make a compelling case for why your compensation should be higher (e.g. cost of living differences), then the recruiter tries to poke holes in the arguments, almost entirely so they can make a strong case to others higher up the chain, and you then hear back if they’re willing to budge (I’m partly guessing here, I don’t know ground truth).
This part of the process seems to have little but not a lot to do with how many papers you’ve published, the no. of citations, etc. in the sense that a candidate with seemingly higher research productivity / quality may end up getting a lower final offer if they don’t negotiate, than someone who may seem less strong on paper but negotiates. There’s significantly more room for numbers to jump around than you might initially expect. And being up against experts at the other end means that they’re likely going to out-convince you if you don’t have experience with this. So candidates shouldn’t self-reject themselves out of the negotiation process and strongly consider companies like Ralph, who can provide candidates with such (potentially unintuitive) guidelines from the start.
We’re excited to have Ralph as a data contributor to add to aipaygrad.es’ publicly available database (with appropriate consent from candidates)! The more data points we have, the more transparent this pipeline becomes, the more informed candidates can be, and hopefully that leads to lesser pay disparity.
— Abhishek Das, Devi Parikh
For compensation data in the AI industry go to www.aipaygrad.es!