Guides
How to choose a machine learning mentor
A good mentor shortens the path by giving you feedback you cannot get from courses: on your code, your experiments and your interviews. A bad one sells a course with a chat attached. Here is how to tell them apart.
What a mentor should do
- Build a plan around your background and goal, not a fixed syllabus for everyone.
- Review your actual code and experiments, with specific comments.
- Meet regularly and keep you accountable.
- Prepare you for interviews with mocks and CV feedback.
- Be honest about timelines and what is realistic.
Match the mentor to the specialisation
Machine learning is broad. A strong computer vision engineer is not automatically the right mentor for quant research, and a researcher may not be the best guide for MLOps. Look for someone who currently works in, or recently worked in, the area you want: ML engineering, LLM and AI engineering, GPU performance or quantitative research.
Red flags
- A job or offer guarantee. Nobody controls hiring decisions.
- No code review: only calls and motivational talk.
- A mass course sold as personal mentorship, with many students per mentor and no individual feedback.
- Unclear pricing, no written contract and no refund terms.
- Pressure to pay immediately after the first call.
Questions to ask on the intro call
- What do you do now, and how does it relate to my goal?
- How exactly is my work reviewed, and how quickly?
- How often do we meet, and what happens between calls?
- What will I have at the end — projects, reports, interview practice?
- What are the payment terms, and what happens if I stop?
Pricing models
Mentoring is usually paid as a fixed price, per session, or with a smaller upfront payment plus a post-payment after you accept an offer. Post-payment aligns incentives but costs more if you succeed; a fixed price is predictable. Whatever the model, insist on a written contract with the calculation, dates and refund conditions.
How to get the most out of mentoring
- Do the practice between calls; mentoring multiplies effort, it does not replace it.
- Bring specific questions and failing experiments, not just progress updates.
- Ask for direct criticism of your code and your project stories.
- Start interview practice early rather than after finishing all the material.
When you do not need a mentor
If you are disciplined, already get feedback at work and have time to search for answers, free resources and communities may be enough. A mentor is most useful when you are switching fields, stuck without feedback, or preparing for a specific set of interviews.
About Quantoros
Quantoros is the personal mentoring practice of Daniil, a quantitative researcher at a financial company whose path went from Python backend through ML engineering to quant research. Mentoring is 1:1: a personal plan, weekly calls, review of every task, mock interviews and CV review, in English or Russian. There are three payment options, including a fixed price without post-payment, terms are fixed in a contract, and there is no job guarantee. The first call is free.
Resources
- Machine Learning Interviews Book — Chip Huyen ↗Understand the hiring process before choosing help.
- Kaggle Learn ↗Free practice if you decide to go solo.
- Stanford CS229 Machine Learning ↗Free university-level ML foundations.
Questions
Is ML mentorship worth it?
It is worth it when you need feedback on real work, a structured plan or interview practice. If you already get these, free resources may be enough.
How much does an ML mentor cost?
It varies widely by format: per session, fixed programmes or upfront plus post-payment. Compare what is included, not only the price.
Should a mentor guarantee a job?
No. A guarantee is a red flag; hiring depends on your work, the market and employers.
Can a quant researcher mentor ML engineers?
Yes, if they have hands-on ML experience. Check the mentor's path and what they review.