Crusoe is an AI infrastructure company founded in 2018 by Chase Lochmiller (CEO) and Cully Cavness (President and COO), with headquarters in Denver and a large engineering presence in the San Francisco Bay Area. It started by capturing wasted flare gas from oil fields to power compute, then rebuilt itself around AI: it now designs and operates data centers, generates and manages the power behind them, and rents NVIDIA and AMD GPU capacity through Crusoe Cloud. The company describes itself as owning every layer from electrons to tokens. It has raised heavily to fund that buildout, including a $600M Series D in early 2025 (about a $2.8B valuation) and a $1.375B Series E in late 2025 (above $10B), with 2026 reports of a further round near a $30B valuation. It also relies on large project financing and joint ventures for data centers, including a multibillion-dollar effort with Blue Owl and Primary Digital Infrastructure on a gigawatt-scale campus in Abilene, Texas.
Process
Public detail on Crusoe’s engineering loop is limited, so treat this as the likely shape rather than a fixed script. Expect a recruiter screen first (background, why AI infrastructure, whether a fast startup fits you), then one or two technical rounds. Glassdoor reports describe the bar as practical rather than LeetCode-heavy, so the coding tends to look like real problems: parse and process data, write clean working code, reason about complexity out loud. A system design round is common and usually maps to Crusoe’s actual product, such as a multi-region control plane, high-throughput low-latency APIs, GPU scheduling, or a service that keeps working when hardware fails. A hiring-manager or founder conversation closes things out, weighted toward ownership and behavioral signal about how you operate with urgency. Candidates on Glassdoor rate difficulty around 2.8 out of 5 with a majority calling the experience positive, and report roughly three weeks from first call to decision.
What they actually ask
- GPU cluster and fleet design: how you would schedule, isolate, and monitor thousands of GPUs across tenants, and what breaks first as you scale that out.
- Networking for AI training: what RDMA, InfiniBand, and RoCE are for, why bandwidth and tail latency matter to distributed training, and how you would debug a slow all-reduce.
- Distributed systems for a cloud control plane: multi-region services, consistency and failure modes, idempotency, and staying available when a region or a rack goes dark.
- Linux and systems fundamentals: processes, memory, and I/O, plus how GPUs and NICs get virtualized and observed, and reading through a performance problem on a real box.
- Backend and API work: designing and scaling internal platforms, provisioning APIs, and the data model behind fleet and customer state.
- Data center, power, and operations for hardware, energy, and SRE roles: capacity and thermal or power limits, automation at scale, what you page on, and a real incident you drove to resolution.
Levels and comp (2026)
Compensation is base plus an annual bonus plus equity, with equity granted as RSUs in a private company on a four-year vest (levels.fyi lists a 25% first-year cliff, then monthly vesting after that). Public data is thin and shifts with each round, so anchor on structure, not a single number. levels.fyi puts software engineer total compensation across a wide band, from roughly $150K at the low end to about $290K near the median, with senior packages reported around $300K and outliers near $500K; staff base salaries in the low-$200Ks plus stock show up in job postings. Because Crusoe is private and raising fast, the equity is the wildcard: RSUs carry real paper value but no liquidity until an exit, and the unit count matters more than the dollar figure a recruiter quotes at today’s preferred-share price. Check levels.fyi and recent Blind threads for current bands, and ask the recruiter for the base, the number of units, the current preferred or 409A price, and the vesting terms so you can value the offer yourself.
Prep priorities
Put most of your time into systems, not algorithm puzzles. Be fluent in distributed-systems basics (consistency, retries, idempotency, sharding, backpressure) and ready to apply them to a cloud control plane or a GPU scheduling service. Know your networking: what RDMA, InfiniBand, and RoCE do, and why they matter to training throughput. Refresh Linux internals and how GPUs and NICs get virtualized and monitored. Have two or three project stories with real trade-offs you owned, since the team screens hard for urgency and end-to-end ownership. If you are targeting a data center, energy, or hardware role, be ready to talk power, cooling, capacity, and automation at gigawatt scale. Write clean, working code under a clock and narrate your reasoning as you go, because the interviewers grade how you think, not just the final answer.
Frequently Asked Questions
Is the Crusoe interview LeetCode-heavy? Reports say no. Glassdoor rates difficulty around 2.8 out of 5, and candidates describe the technical rounds as practical: real problems, clean code, and clear reasoning rather than obscure algorithm tricks. System design tied to their infrastructure carries a lot of the weight.
How long does the process take? Glassdoor puts the average near three weeks from first contact to offer. A startup at this stage can move faster when the team likes a candidate, so keep your scheduling flexible and follow up quickly.
Do I need GPU or data center experience to get hired? Not always. Backend, distributed systems, and platform engineers are hired across Crusoe Cloud. Depth in networking, GPUs, or power helps for systems and infrastructure roles, but strong fundamentals and evidence you can own a system end to end matter most.
What does Crusoe want beyond technical skill? Urgency and ownership. The company says it wants teammates who find problems and move fast where the path is not fully paved, so bring examples of driving work through ambiguity rather than only executing well-defined tasks.
Similar company guides
Prepping for Crusoe? Put it to work:
