I've now been through the Jane Street process twice: once in 2023 (rejected at phone screen) and once in 2025 (offer, declined). Here's what I'd do differently starting from zero.
First thing to understand: this is not a Leetcode grind company. Yes, there's coding, and yes, it needs to be good, but the path to passing isn't hammering Leetcode mediums. The path is being genuinely fluent in functional programming, having solid probability intuition, and being able to reason clearly under pressure.
OCaml first, before anything else. They use OCaml internally and the coding rounds are in OCaml. Get comfortable enough to write idiomatic OCaml quickly. I'd suggest actually building a small project, not just doing exercises. Real-Ocaml.org is solid. Budget 6-8 weeks if you're coming from Python or Java.
Probability and combinatorics. Work through the Jane Street puzzle archive, they post them publicly. Also work through classic quant interview books like Heard on the Street or anything by Mark Joshi. The expected value problems and Bayesian reasoning questions are real and they matter.
Simulate the trading round specifically. Find someone to roleplay market-making with you. Practice quoting bid/ask spreads on made-up problems. Explain why you'd move your price given new information. This sounds weird but it's a specific skill and it's learnable.
Think out loud constantly. Every problem, every day, narrate what you're doing. Jane Street interviews are as much about how you reason as what you arrive at. Interviewers are evaluating your thinking process.
Behavioral prep. Don't skip this. They do ask it, and they want specific examples of how you've handled ambiguous problems. The typical STAR format works here.
The loop is long and genuinely hard. But it's also one of the most intellectually interesting interview processes I've done. Worth the effort even if you end up elsewhere.