Skild AI Interview Guide (2026): Robot-Brain Research and Engineering Rounds

Skild AI is a robotics startup founded in 2023 by Deepak Pathak (CEO) and Abhinav Gupta (President), both former Carnegie Mellon professors. It is building the “Skild Brain,” a single foundation model designed to control many robot types (quadrupeds, humanoids, tabletop arms, and mobile manipulators) rather than one product robot. The company runs offices in Pittsburgh and the San Francisco Bay Area, with additional hiring in Bengaluru. In January 2026 it raised a reported $1.4 billion round at a valuation above $14 billion, led by SoftBank with backers including Nvidia’s venture arm, Lightspeed, Coatue, and Jeff Bezos, following a $300 million round in 2024. Skild’s thesis is that scale in data and learning, rather than hardware built for one task, is what lets robots handle unseen situations, and it describes the Skild Brain as omni-bodied, meaning one model that adapts to a robot without prior knowledge of its exact form. That thesis and the frontier-lab talent bar shape how the company hires.

Process

Public detail on Skild’s loop is limited, and the few candidate accounts do not fully agree, so treat this as the likely shape rather than a fixed script. Expect a recruiter screen first (background, why robotics, why now), then a split by track. For engineering roles (robotics software, AI inference, training infrastructure), plan for one or two technical rounds heavy on strong coding plus ML system design tied to data pipelines and model serving. For research roles (Research Scientist and Research Engineer), expect a presentation of your past work plus discussion of reinforcement learning, imitation learning, and generalization, since the posted bar is a PhD with publications at venues like NeurIPS, ICML, CoRL, RSS, and ICRA. Most loops close with a team or founder conversation weighted toward ownership and behavioral signal. Cycle length is hard to pin down: plan for several weeks, and note that some candidates report slow or unclear recruiter communication and long gaps, so ask for a timeline up front. Hardware and embedded roles skew more toward your resume and past systems work, and one candidate account described a round focused mostly on prior projects with weight on electrical engineering detail, so match your prep to the exact role.

What they actually ask

• Strong coding under time pressure: core data structures and algorithms with a clean, working implementation. For infrastructure roles, expect systems-level questions on performance, memory, and concurrency.
• Robot-learning depth: how you would train a generalist policy with imitation learning and reinforcement learning, and how you handle many robot embodiments from one model.
• Sim-to-real: closing the gap between simulation and hardware, domain randomization, and why a policy that works in sim can fail on a real robot.
• Robot data pipelines: collecting, cleaning, and scaling teleoperation and real-world robot data, and how data quality changes model behavior.
• Distributed training: training large models across many GPUs, handling throughput and checkpointing, and debugging a run that diverges partway through.
• Research depth (for scientists): a walk-through of a paper or project you led, with pushback on your methods, baselines, and what you would do differently now.

Levels and comp (2026)

Skild has not published pay bands, and it is a private, freshly funded company, so treat any number as approximate. Compensation is base salary plus equity plus benefits, with the equity typically on a four-year vest and a one-year cliff at this stage. Job-board and aggregator listings put base salaries across a wide span, roughly from the low six figures up to around $300K depending on level and role, with senior engineering near the top of that and research-scientist offers set to compete with frontier AI labs given the PhD bar. Equity is the real variable: at a valuation above $14 billion the paper number can look large, but its worth depends entirely on the outcome and the price you come in at, so ask for the base, the number of units, the current preferred-share price, and the vesting terms so you can value it yourself. Bands move fast right after a big round, so check levels.fyi and recent Blind threads for current numbers before you anchor on any figure.

Prep priorities

Put real time into strong coding under a clock, since fast, clean solutions matter across every engineering track. For research and ML roles, prepare a tight talk on your best robot-learning or ML work with clear baselines, ablations, and results, and expect hard questions on your methods. Refresh reinforcement learning and imitation learning fundamentals, sim-to-real techniques, and how large policies get trained and served. On the infrastructure side, review distributed training, data pipelines, and serving latency. Have two or three project stories with real trade-offs you owned, and read Skild’s own writing on the Skild Brain so you can talk about generalist, omni-bodied policies in their terms. If a round is a research presentation, rehearse it end to end and prepare for questions that probe why your approach works.

Frequently Asked Questions

Does Skild AI require a PhD? For Research Scientist roles the posted bar is a PhD plus publications at top venues such as NeurIPS, ICML, CoRL, and ICRA. Many engineering roles (robotics software, AI inference, infrastructure) do not require one and instead value shipped systems and strong coding.

Is the interview more about research or engineering? It depends on your track. Research roles center on your past work and robot-learning depth, while engineering roles lean on coding, systems, and practical robotics. Ask the recruiter which loop you are in so you prep the right things.

How long does the process take? Plan for several weeks. Some candidates report slow recruiter communication and long gaps between stages, so confirm the timeline and next steps at each step rather than assuming a fast turnaround.

Does Skild sponsor visas? Reports indicate it does sponsor for many roles, but confirm with the recruiter for your specific role and location before you count on it.