Castelion builds hypersonic strike weapons and is trying to do it the way SpaceX did rockets: vertical integration, rapid flight-test iteration, and manufacturing designed for rate and unit cost from day one. Its first system, Blackbeard, is pitched as the first U.S. hypersonic weapon engineered from the start for industrial-rate output and commercial-style unit economics. The founders came out of SpaceX, and the company has said it is chasing affordable, mass-produced deterrence rather than the small batches of exquisite, expensive missiles the primes have historically delivered.
The money and the contracts are what make this loop worth preparing for now. Castelion closed a $350 million Series B to scale hypersonic production, and, according to reporting, by August 2026 it had raised a roughly $1 billion Series C at about a $13 billion valuation, led by a16z with Carlyle and JPMorgan participating, on the back of a reported $500 million-plus in U.S. military contracts. That surge in capital and demand is why it is hiring across avionics, flight sciences, guidance-navigation-and-control, materials, mechanical, radar, software, and manufacturing at the same time. It rides the same 2026 defense-tech wave as Anduril, Saronic, and True Anomaly, and if you are weighing several of these bets, the broader set of company interview guides and the peer True Anomaly guide put the stage and the culture in context.
Why the questions lean toward first principles
Read the product as an engineering brief and the interview follows from it. A hypersonic vehicle has to survive extreme aerothermal loads, hit a guidance solution under conditions you cannot fully test on the ground, and, uniquely here, come off a production line at rate and at a price the government will actually buy in quantity. That last constraint is the whole thesis. So interviewers care less about whether you have memorized an algorithm and more about whether you can reason from physics and cost toward a design that manufactures.
For hardware roles that means thermodynamics, structures, aerodynamics, or GNC reasoning done out loud, usually anchored to something you actually built. For software it means embedded and flight-adjacent work, real-time systems, hardware-in-the-loop testing, ground software, and simulation, not web CRUD. For manufacturing it means design-for-manufacturing, process development, tolerancing, and how you take a prototype to rate production without a year of hand-tuning. The common thread is that Castelion, like SpaceX, rewards people who take ownership of an ambiguous problem and move fast, and it filters hard against people who need a fully specified ticket before they start.
The interview loop, round by round
The structure below is inferred from how SpaceX, Anduril, and True Anomaly run their hardware and software loops, plus Castelion’s own public job postings; the company has not published its process, so treat this as a well-informed sketch rather than a leaked script. Expect roughly four to six touchpoints, weighted toward the team you would join, and expect the loop to move quickly given how fast the company is scaling.
| Castelion interview stage | Format and length | Who runs it | What it screens for |
|---|---|---|---|
| Recruiter screen | ~30 min phone | Recruiter | U.S.-person/ITAR eligibility, clearance eligibility, mission motivation, on-site willingness, resume walk |
| Hiring-manager technical call | ~45-60 min video | Team lead you would report to | Domain depth, how you reason about a real problem, ownership of past projects |
| Discipline deep dive | On-site panel, half to full day | Engineers on the team | First-principles design and tradeoffs; coding for software, design review for hardware, DFM and process for manufacturing |
| Practical exercise | Whiteboard or short take-home | Team | Concrete problem-solving: a coding problem, a back-of-envelope sizing, or a manufacturing process plan |
| Mission and ownership | ~45 min | Founder or senior leader | Why defense, why now, bias to ship, comfort with pace and ambiguity |
The deep-dive round is where most candidates pass or fail, and it is discipline-specific. Software candidates should expect practical coding at a data-structures level plus embedded and systems questions; the bar is clean, correct code and clear reasoning about real-time behavior, not exotic dynamic-programming puzzles. Reviewing core coding patterns is enough coverage for the algorithmic slice, and a short study plan keeps that from eating your whole prep budget, because the flight-software and hardware-in-the-loop conversation matters more here than raw LeetCode.
Hardware candidates get a design review instead of a coding screen: walk us through something you built, why you sized it that way, what you would change, and what broke. Manufacturing candidates should expect to reason about taking a part from prototype to rate, tolerancing, tooling, and where a design fights the process. Across all three, someone will push on tradeoffs until you either defend your reasoning or revise it, and revising it under pressure is a good sign, not a bad one.
The screen that has nothing to do with engineering
Before any of the technical rounds matter, there is a hard eligibility gate, and it is worth understanding early because it will save you a wasted loop. There are two separate gates here, and conflating them is what trips people up. The first is ITAR “U.S. person” status, which governs eligibility to touch the export-controlled work at all: a U.S. citizen, U.S. national, lawful permanent resident (green-card holder), refugee, or asylee qualifies, and this is the bar that applies to almost every role. The second, narrower gate is a U.S. government security clearance. Many Castelion postings require the ability to obtain and maintain a Top Secret/SCI clearance, and that subset needs U.S. citizenship specifically, because green-card holders and other non-citizens cannot be granted TS/SCI. So a lawful permanent resident is generally eligible for the ITAR work but not for the clearance-gated roles. If you are on a visa and would need sponsorship, though, this is not a company that can hire you right now regardless of how strong your engineering is, and the recruiter screen exists partly to confirm that fast.
The other non-negotiable is location and pace. Roles are concentrated in the Los Angeles area, around Torrance and El Segundo, and are on-site by default, with some flight-test roles adding travel to government ranges. This is a build-hardware-in-a-factory company, not a remote one, so if relocation to Southern California or long hours near a production line are dealbreakers, sort that out before you invest in the loop rather than after.
How to read the comp
There is no reliable public levels.fyi figure for Castelion yet; the page explicitly says it needs more submissions to display one, so anyone quoting you a precise Castelion total-comp number is guessing. What you can anchor on is the surrounding market. Ex-SpaceX defense-hardware startups in high-cost California tend to pay competitive base salaries with meaningful early-stage equity and lighter cash bonuses than big tech, and California pay-transparency law means the actual posted ranges are on the job listings themselves, which is the most reliable source you have. Read the range on the specific req you want, not a blended average.
As a mental model, expect base bands broadly in line with SpaceX and Anduril for the same level and discipline, with the equity carrying more of the upside given a reported valuation, roughly $13 billion, that has climbed fast. For senior hardware and software engineers, that tends to mean base salaries roughly in the $150k to $250k area depending on level and discipline, with the real number set by the posted California range on the specific req rather than any figure quoted here. Treat that band as a rough anchor to verify against the listing, not a promise. Run any offer through a total-comp calculator so the equity is valued realistically against the current valuation and your expected vesting, and when the number comes back, the mechanics in a good salary-negotiation walkthrough apply here as much as anywhere. Early-stage equity at a fast-appreciating private company is the part most people misjudge, in both directions.
How to prepare
Come in fluent in the mission and specific about your own work. Castelion screens hard for people who actually want to build weapons for national security and can say why without either flinching or sloganeering, so have a real answer to why defense and why now. Then bring one or two projects you can go deep on for forty-five minutes: what you built, the constraints, the tradeoffs you made, the numbers, and what you would do differently. Vague ownership claims fall apart fast in the deep dive. Postings skew toward experienced engineers, but expect the bar to favor demonstrated hands-on building, so a substantial capstone, a rocketry or Formula SAE club project, or a serious internship where you owned real hardware or flight software can count as the thing you built, provided you can defend the decisions the way a senior would.
For software, drill practical coding and shore up embedded, real-time, and simulation fundamentals; for hardware, be ready to size and defend a design from first principles; for manufacturing, be ready to talk rate production and DFM in concrete terms. The behavioral round is a genuine filter on pace and ownership, not a formality, so prepare a couple of tight stories the way you would for any high-bar loop. If your ownership stories tend to ramble, the STAR-method structure keeps them to the point, and running your resume through a resume checker first is worth the ten minutes because a hardware-heavy resume with quantified results reads very differently from a generic software one.
The thing to internalize is that Castelion is not hiring you to optimize a known system. It is hiring you to build one that does not exist yet, at a rate nobody in this category has hit, under real security constraints, and it interviews accordingly. If that sounds like the best job you could have or the worst, the loop is designed to tell you which one it is.
Practice the behavioral round:
