Leveling & How Decisions Are Made

An Optiver offer sits inside a small firm, roughly 2,100 people with under a thousand engineers, and that scale shapes how leveling works. There is no sprawling grade ladder with a rung for every year of tenure. A software role lands on one of three public titles, Graduate Software Engineer, then Software Engineer, then Senior Software Engineer, with hardware, FPGA, and the quant tracks running as separate ladders alongside. Settle which title the requisition is really for before the loop starts, because at a firm this size the band travels with the posting more than it moves during the interviews.

The Optiver angle

The bar you have to clear is not the trader's math test. Optiver's public reputation is the timed mental-arithmetic sprint, eighty questions in eight minutes with no calculator, and the market-making games on a final day. That reputation belongs to the Trader and Quant loop. The Software Engineer loop is a different animal: a HackerRank coding assessment, an early behavioural conversation, a technical interview, and a final day of engineering rounds. Optiver's own published loops keep the two apart, so read every Optiver interview story for which role it belongs to before you borrow its prep. An engineer who grinds arithmetic drills is training for a gate that never appears.

The Software Engineer Ladder

Optiver does not publish a detailed internal grade rubric, so the ladder is best read through the three public titles and the scope each one owns. Graduate Software Engineer is the entry rung for campus and new-grad hires. Every graduate starts at the Global Optiver Academy in Amsterdam, including those bound for the US and APAC offices. It is a structured onboarding program of roughly ten to twelve weeks before the engineer joins a team. Software Engineer is the standard experienced-hire title, owning features and systems inside a team. Senior Software Engineer carries broader technical ownership. There is no public Staff or Principal title on the engineering side. Where each rung sits is driven by demonstrated scope rather than a fixed tenure gate, and Optiver keeps the internal calibration for that call private. Compensation is not covered here; it lives in the negotiation lesson.

• Graduate Software Engineer: entry / new-grad. The campus and new-grad rung. Every graduate, US and APAC included, starts through the Global Optiver Academy in Amsterdam, roughly ten to twelve weeks of onboarding, before joining a team. levels.fyi lines this up with L3.
• Software Engineer: experienced hire. The standard mid-level title. You own features and systems inside a team and write low-latency C++ against live trading infrastructure. levels.fyi maps the band to L4 and L5.
• Senior Software Engineer: senior IC. Broader technical ownership across systems, and in some teams technical direction. levels.fyi maps this to L6. Sparsely documented publicly.
• Separate tracks: FPGA / hardware / quant. FPGA and hardware engineering run their own ladders with Verilog and VHDL rounds, and the Quant Trader, Researcher, and Risk tracks are distinct again. Different loop, different bar.

Who Decides, and How

The Software Engineer pipeline runs in a clear order. Application, online assessment, an early behavioural interview, then a technical interview. Last comes a final interview day of two technical rounds and one behavioural. Optiver publishes that shape in its own recruiting FAQ, so the sequence is solid. The exact round count above that minimum, and who staffs each seat, shifts by office and cycle. In the US offices the final day is run virtually rather than as an in-person fly-out. The recruiter coordinates the loop and carries the outcome, and the engineers on the panel own the technical read. How those reads convert into a single yes or no is not something Optiver publishes. There is no public sign of an Amazon-style dedicated bar raiser or a cross-org hiring committee. No absence is confirmed either. Treat the decision as panel and hiring-team led, and hold the finer mechanics loosely. Early stages give no individual feedback, and a rejection can be headcount-driven rather than a verdict on your ability.

The Optiver angle

A strong assessment score is a ticket to the room, not a pass. The engineering bar here is holistic and genuinely steep. Some engineers have scored around ninety percent on the online assessment and still been turned down. The assessment grades correctness against hidden tests. The decision weighs things harder to score. Does your object-oriented model stay clean under a dense spec? Do you reason about memory and latency without being asked? Can you defend a design choice when it is pushed? Clearing the coding gate proves you can code. It does not prove the depth the panel is there to find, and that gap is where capable engineers are most often stopped.

What the Final Day Weighs

The final day is where the bar shows its shape. Optiver's own guidance for engineering candidates names what the technical rounds test. You need command of one object-oriented language and its standard library, and data structures with honest time and space reasoning. You need to know how those structures sit in memory and how operations work underneath. Add literacy across computer architecture, networking, concurrency and memory management, all pointed at low-latency systems thinking. The firm states plainly that these are not standard algorithm puzzles or brainteasers, and the rounds are collaborative. You are expected to think aloud and defend your decisions rather than recite a memorized optimum. Running alongside the technical rounds is the behavioural interview, which appears early in the loop and again on the final day. It uses the STAR method and is openly two-way, probing motivation, collaboration, and fit while leaving you room to judge whether Optiver is right for you. The near-universal opener is some form of why Optiver, or why trading, so come with a specific answer.

The Optiver angle

What separates this from a FAANG loop is systems literacy. Picture a candidate who can invert a binary tree in three languages but goes quiet on how a hash map is laid out in memory. Or on why a cache miss costs what it does. For a firm whose product is speed, that reads as underpowered. The engineers on the panel build systems where the path from market data in to orders out is measured in nanoseconds, and they weigh a final-day borderline on whether you think in those terms. Show that you reason about the machine, defend your tradeoffs, and write code someone else can maintain, and you clear a bar that a pure puzzle specialist does not.

Common questions