Preferred Networks Reviews

PFN is an unsusual company. In the context of Japanese tech companies, there are few that can rival PFN in terms of the experience and brilliance of the individual employees in areas of machine learning and software support for ML. Unfortunately the high-level (and maybe especially mid-level) management is woefully inexperienced in areas of management and business. It's unclear exactly why this situation was created or allowed to persist for years, but it may have to do with two issues: hero/engineer worship, and cronyism. To the first, PFN is an extremely engineer-driven company. On one hand, this is great. The engineers you work with? They're talented engineers. But your manager? He's an engineer. His manager? Also an engineer. The web designer? You guessed it: engineer. The business devs? Many are engineers. There seems to be a deep-seated view that an engineer can do anything. Unfortunately, they often will have to do anything. This is often phrased as a company value - "Learn or Die" - it is literally the title of a book about PFN. But while the aspiration to adapt and transform yourself into these varying roles is admirable, ultimately this creates managers, designers, and business devs, etc. with functionally 0 experience. And what an insult it is to the people whose career and education are in one of these professions, to think that a high coding score and some nights spent reading about the role online would outweigh years of education and experience. And ironically, despite the typical employee being often especially good at a specialized task, it seems that their opinions, both in their area of expertise and in general, are often not considered in any decision making process. Having many people in roles they're not suited for has likely created innumerable problems, but notably among them: when basic essential skills for a job are no longer required to do it, promotion to those positions need no longer be tied to actual ability to perform it. Meaning that the promotion requirements appear to be quite arbitrary and tied more to how long you've known everyone than any performance metric. You would never be able to say that some business success caused some promotion, and simultaneously, business failures are never met with demotion. And when this is the only career path offered, to most employees, it is a dead end. Another review commented on a lack of transparency. This may be artfully underselling it -- transparency is not an endangered species at PFN, it is an extinct one. As described above, there is little transparency in promotion. There is also little transparency in how projects are cancelled. And probably most important to the average employee, there is little transparency in evaluation or compensation. The latter has created a somewhat toxic environment. Similar performance can be met with wildly different evaluations depending on which team someone is on, and which peers will be evaluating them. There is no objective, clear, universal, or agreed-in-advance metric for evaluation and employees, and no oversight. It can be decided largely on the whim of a single person, with a justification not really any more meaningful than "Because I said so". That is not to say that there are not pockets within PFN where a pleasant work life is possible. PFN is not a black company, and you will never be forced to work until 3am unpaid, or be guilted into never taking your holidays. While many Tokyo-based tech companies now pay significantly more than PFN, it is still a reasonable option, especially for those wanting a strongly ML-focused job. However, for talented candidates who want to progress their career, to be given opportunities to shine and to be acknowledged and rewarded for doing it, it could easily turn into a 1-star experience. So to potential employees, I suggest to ask in interviews: - who is my manager and what is his experience? - who is their manager and what is their experience? - what team will I be working on? - what will happen to me if my project is cut? - what are my options for career growth (be specific)? - what is my raise and bonus structure for the next n years? If you do not receive a clear answer to these questions, it is very likely you will not like the answer when you discover it on the job.