Human Computer’s median salary band hits $178,000 for hybrid defense-tech roles

The Roles That Map the Vector

Human Computer has six active roles on its board, four with published salary bands, and the median band hits $178,000 — a figure that aligns with defense-prime principal-engineer compensation, not consumer-tech benchmarks. The highest-band role, a Senior Product Manager for Client Side Defense in New York, sits at the intersection of product and defense procurement. A Staff Product Manager (U.S. remote) follows. Both titles signal the firm is building product muscle around defense customers — shaping roadmaps that survive classification reviews, procurement cycles, and the particular rigor of "client-side" delivery where the end user wears a uniform or holds a clearance.

The security research tier runs parallel. A Senior Security Researcher (U.S. remote) lists, while a Threat Research Analyst (U.S. remote) sits. The gap is deliberate: the senior role expects original vulnerability discovery, tooling development, and likely briefing stakeholders who operate in SCIFs. The analyst role skews operational: triage, enrichment, feeding detection pipelines. Together they form a classic research-to-operations funnel that only pays off when the team shares context with the product managers defining what gets built.

Tel Aviv hosts two listings without published bands: a UX/UI Designer and a People Operations Specialist covering maternity leave. The designer role matters disproportionately. Defense and security products that fail usability tests don't just annoy users — they get worked around, and workarounds in classified environments become security incidents. Placing design in Tel Aviv, alongside a people-ops hire, suggests Human Computer treats its Israel office as a full-stack product center, not a back-office satellite.

What's absent is equally telling. No pure ML research roles. No embedded firmware listings. No hardware bring-up titles. The hiring plan leans hard on the software-to-mission translation layer: product managers who speak acquisition language, researchers who output actionable intel, designers who harden interfaces against operator error. That's the hybrid profile the broader market is scrambling to define; the six visible openings form a coherent thesis: Human Computer is staffing the seam where AI-enabled systems meet defense procurement reality.

The Screen Is a Compatibility Check, Not a Skills Test

The recruiter screen functions as the first real filter in frontier tech hiring — a 30-minute call that operates less like a skills test and more like an initial compatibility check. As one candidate who completed roughly 30 such screens described it, the conversation is "kind of like that initial chat that you have with this person that you've matched with" on a dating app. The recruiter evaluates baseline fit; the candidate evaluates whether the company warrants further time investment. Most screens are run by a dedicated recruiter, though hiring managers or directors sometimes lead them directly. The structure varies wildly: some companies run one technical phone screen, others run two, and on-site loops can span five sessions across two days.

Human Computer's open roles, spanning defense-facing product management, security research, and UX for AI-integrated systems, sit inside that $178,000 median band, signaling the company is hiring for senior IC and lead-level talent, not junior generalists. The screen for these roles rarely involves technical trivia. The same candidate who logged 30 recruiter calls encountered technical questions ("tell me about the difference between var, let, and const in JavaScript") exactly once. The norm is broader: can you articulate what you've built, why it mattered, and whether you understand the problem space the role sits in?

Frontier firms screening for AI-integrated systems, robotics, and defense-adjacent work tend to weight three non-traditional signals heavier than pedigree. First, domain fluency demonstrated through context — not a certificate or course completion, but the ability to explain a technical choice in terms of the physical or operational constraints it addressed. A security researcher who can describe how a side-channel mitigation interacted with firmware update cycles on deployed hardware passes further than one who lists the CVE number. Second, systems-level ownership — evidence that you've operated across the software-hardware boundary, owned a subsystem from spec to field deployment, or debugged a failure that spanned simulation and silicon. Third, bidirectional evaluation discipline — candidates who treat the screen as a two-way process, asking pointed questions about team structure, technical debt, mission alignment, and red flags, signal the maturity these roles demand. The same candidate who survived 30 screens emphasized: "use this opportunity to also look for red flags you don't want to end up wasting your time interviewing with this company if you think they're not going to be respectful of your time."

What disqualifies candidates at this stage is rarely a knowledge gap. It's an inability to translate experience into the language of the problem set. Recruiters at frontier firms are trained to listen for specificity: not "I optimized the pipeline," but "I cut inference latency 40% by rewriting the batching logic to match the accelerator's memory bandwidth profile." The screen tests whether you can speak the dialect of the domain; the next round, the technical deep-dive, assumes you already do.

Pedigree Signals Access; Fluency Signals Survival

The resume made sense when careers followed a single craft for decades. That model collapsed. Vervo's founder put it plainly: the document is "essentially like a career chronology… that made complete sense when our careers were organized in that way, like 400 years ago when you had to do a blacksmith apprenticeship for 10 years." Today the best performers at frontier companies aren't the ones with the most prestigious resumes. "They were the curious, resourceful, and tenacious ones. And none of that showed up on paper."

Human Computer's open roles illustrate the shift. These aren't checkbox positions. They demand fluency in threat modeling, defense procurement cycles, and the interplay between software controls and physical effects, knowledge you don't absorb from a brand-name employer's onboarding program.

"Analogous thinking actually becomes even more important because you can draw on different experiences to help you make better decisions and it becomes less important that you did the same craft for 20 years when that craft itself could become obsolete."

AI commoditized knowledge retrieval. A 19-year-old intern can now reach expert-level information in hours. What remains scarce is the judgment to apply that information inside a specific system: a satellite ground station, a robotic manipulator, a classified network. Pedigree signals access to a curriculum. Domain fluency signals survival in the environment where the curriculum gets tested.

Companies clinging to the old measurement problem filter for resume aesthetics. Vervo's data shows they miss "hidden gems that actually going to be great performers but for whatever reason they don't stand out from the pile." The alternative is task-based assessment: give candidates a realistic slice of the work (a threat-hunt scenario, a requirements trade-off for a defense payload) and evaluate the output. "Make the assessment task based and practical rather than knowledge based cuz if it's knowledge based you're asking to be lied to."

The old guard resists. Some hiring managers still insist candidates write resumes manually while expecting them to use AI on the job. That contradiction reveals the real barrier: not talent scarcity, but measurement inertia. Frontier employers who switch to practical evaluation find the pipeline widens. The ones who don't keep hiring the same pedigree profiles and wondering why the integration failures persist.

Hybrid Roles Are the Only Kind That Scale

Robotics talent is hybrid by nature. Software meets hardware. AI meets control systems. Simulation meets real-world failure modes. The supply side does not scale cleanly because the skill stack does not form quickly. You cannot shortcut years of systems thinking, field exposure, and cross-domain depth. Training pipelines lag. Universities trail industry reality. Experience remains scarce by definition.

The convergence driving this scarcity is not theoretical. Manufacturers chase throughput. Logistics chases precision. Healthcare chases reliability. Defense and aerospace chase autonomy at scale. Each sector pulls from the same finite group of engineers who understand perception, motion planning, embedded systems, and AI inference in physical environments. Demand multiplies horizontally while supply inches forward.

The market signal is unambiguous. Job postings for "Robot Learning Engineer" (the single hottest role in 2026) have grown 340% since January 2024, dwarfing growth in every other robotics specialization. These specialists train manipulation and locomotion policies using imitation learning and foundation models, commanding senior compensation packages when equity is included. Their day-to-day work spans designing and training policies (behavioral cloning, diffusion policies, ACT), managing large-scale data pipelines ingesting teleoperation demonstrations, sim-to-real transfer through Isaac Sim or MuJoCo, and systematic real-world policy evaluation.

A strong robotics engineer is rarely a narrow specialist. Even roles labeled as "software" require comfort across sensors and perception, embedded systems and real-time constraints, control theory and motion planning, physical system limitations, and failure modes and safety considerations.

The skill stack reflects this breadth. Core ML demands PyTorch, transformer architectures, diffusion models, and VLA (Vision-Language-Action) model fine-tuning. The robotics stack requires ROS 2, URDF/MJCF, real-time control at 100+ Hz, and sensor fusion across RGBD, force-torque, and proprioception. Data engineering means large-scale dataset management in RLDS format, data quality scoring, and replay tools. Simulation fluency covers NVIDIA Isaac Sim, MuJoCo, and domain randomization. Hardware fluency means the ability to work directly with physical robots, debug hardware-software integration issues, and iterate in the real world.

Companies building foundation models for robot manipulation (Physical Intelligence, Skild AI, Google DeepMind Robotics, NVIDIA) compete fiercely for engineers who bridge VLA architectures, sim-to-real transfer, and large-scale demonstration data pipelines. As data collection becomes the bottleneck for scaling robot learning, engineers who can build reliable, low-latency teleoperation infrastructure are in rapidly increasing demand. Teleoperation System Engineers, while compensated below the median, represent one of the fastest-growing categories.

This hybrid demand extends beyond robotics proper. Human Computer's current openings sit at the intersection of AI, defense systems, and product execution. The board's salary band reflects the premium for talent that translates between technical depth and operational deployment. Defense and energy sectors increasingly require engineers who can integrate AI inference into physical platforms with safety-critical constraints, not just cloud environments.

In 2025, hiring priorities shifted decisively toward mechatronics, embedded systems, and hands-on integration experience over narrow software expertise alone. Writing clean code is expected. Understanding how that code behaves when motors stall, sensors drift, or latency spikes is what differentiates strong candidates. Mistakes in this domain are not abstract — robots move, carry loads, operate near people. Engineers working on industrial robots, autonomous mobile robots, or collaborative systems must understand safety standards, fault tolerance, and compliance requirements. This experience cannot be inferred from algorithm tests or generic coding challenges. It must be demonstrated.

The talent gap is structural. Over 4.2 million AI-related roles are projected worldwide against barely 2.1 million qualified professionals. That gap does not narrow. It hardens. Robotics and AI roles in manufacturing and industrial environments sit open for more than five months on average. Germany alone has faced over 320,000 unfilled STEM roles in recent years. By 2029, the US is expected to have approximately 172,300 robotics engineering roles, nearly 9% higher than 2024 and more than 40% above the 2013 low point.

Every robotics roadmap today is gated by people, not ambition. The capital is ready. The use cases are proven. The demand signal is loud. What breaks the system is execution, and execution breaks at hiring.

The Pipeline Is Rewiring Around Demonstrated Capability

The hiring data from Human Computer reflects a pattern that university career offices and bootcamp operators are only beginning to recognize. Traditional signaling mechanisms (elite degrees, brand-name internships, GPA thresholds) are losing predictive value for the hybrid roles now appearing on frontier-tech boards. The board data shows Human Computer recruiting across product, security research, threat analysis, and design, with locations split between New York and Tel Aviv. None of these listings mention degree requirements. The median band of $178,000 suggests these are not entry-level positions, yet the absence of credential filters signals a shift toward demonstrated capability over institutional validation.

Universities face a structural lag. Curriculum committees operate on multi-year cycles; a new course in AI-integrated systems engineering takes 18 months to approve, by which point the stack has moved. Computer science departments still graduate students who have never touched a real-time control loop, never debugged a perception pipeline on hardware, never written a requirements document for a system that must survive vibration, thermal cycling, and adversarial input. The Human Computer roles (particularly the Senior Security Researcher and Threat Research Analyst positions) demand fluency in threat modeling for deployed systems, not classroom exercises. That gap cannot be closed by adding an elective. It requires rethinking the capstone experience, the industry partnership model, and the faculty incentive structure that rewards publication over deployment.

Bootcamps have a different problem. Their 12-to-24-week format forces compression, which works for syntax and framework familiarity but fails at systems depth. A graduate who can spin up a LangChain prototype in a weekend cannot necessarily reason about latency budgets when that prototype runs on an edge device with 4 GB of RAM and a 50 ms cycle time. The Staff Product Manager role at Human Computer implies ownership of technical trade-offs across software, hardware, and operational constraints — exactly the synthesis that short programs cannot teach. Some bootcamps have responded by adding "systems engineering" modules, but without access to physical testbeds (flight hardware, robotic arms, radiation-hardened compute) those modules remain theoretical.

Self-directed learners occupy the most volatile position. The open-source contribution record has become a de facto portfolio for frontier roles, but the signal is noisy. A GitHub history full of cloned tutorials and low-effort PRs reads differently than a single well-documented contribution to a real-time OS scheduler or a verified exploit write-up for an embedded modem. The Human Computer security research roles will filter for the latter. Candidates who understand this are building in public — publishing postmortems of failed hardware bring-ups, releasing tooling for side-channel analysis, documenting the process of porting a perception stack to a new compute module. These artifacts carry more weight than a master's thesis because they prove the candidate has wrestled with the same constraints the hiring team faces daily.

The pipeline implication is not that credentials are dead. Security clearances still require citizenship and background investigation. Export-controlled work still demands ITAR compliance training. But the technical filter has moved upstream. Universities that embed their students in live programs (not senior design projects sponsored by a company logo, but actual subcontracts with deliverables) will place graduates. Bootcamps that partner with hardware labs for capstone access will differentiate. Self-directed learners who treat their portfolio as a body of evidence, not a highlight reel, will pass screens. The rest will keep optimizing for a rubric the market has already discarded.

This Signal Is Narrow, Not Universal

Human Computer's current slate (four salaried roles on the board, one added in the past seven days) is a signal, not a surge. The positions cluster in defense-adjacent product and security: the senior product manager role in New York, the staff product manager role (U.S.), the senior security researcher role (U.S.), the threat research analyst role (U.S.), plus the designer and people operations roles in Tel Aviv. The board's median salary band sits at $178,000 across a $104,000–$218,000 range. That is a specialized hiring footprint, not a broad labor-market indicator.

Extrapolating a macro boom from a half-dozen openings at one firm commits a category error. The U.S. tech labor market shed over 260,000 jobs in 2023 alone, and while frontier subsectors (defense tech, AI infrastructure, robotics) have held tighter, they represent a fraction of total engineering employment. Human Computer's demand reflects a specific procurement cycle: defense customers buying AI-integrated tooling, and the company staffing to deliver. That cycle is real. It is also narrow. A single company's product and security hiring does not equal a sector-wide recovery, let alone a general tech rebound.

The roles themselves underscore the sector specificity. Client Side Defense product management is not interchangeable with consumer SaaS product management. Threat research for defense-adjacent systems demands clearance eligibility, air-gapped environment experience, and familiarity with adversary TTPs that don't transfer from commercial SOC work. The Senior Security Researcher role leans into vulnerability research and exploit development — skill sets that remain scarce but are concentrated in a handful of contractors, FFRDCs, and a few commercial shops. These are not generic "AI engineer" requisitions. They are domain-locked.

Nor does the screening shift described in earlier sections amount to a rejection of fundamentals. Human Computer's listings still require deep technical fluency: the security researcher role expects C/C++, reverse engineering, and kernel internals; the product roles demand experience shipping to government customers with compliance overhead (FedRAMP, ITAR, CMMC). The "hybrid" profile (software fluency plus hardware or domain context) is an addition to the baseline, not a substitution. Candidates who cannot pass a systems design review or a kernel debugging exercise do not advance, regardless of their portfolio projects. The fundamentals gate remains; the domain gate has widened.

The geographic concentration further limits generalization. Two roles sit in Tel Aviv, three in the United States (one explicitly New York). This reflects Human Computer's existing footprint and its customer base (Israeli defense establishment, U.S. prime contractors) not a distributed frontier-tech hiring wave. Other frontier employers (Anduril, Shield AI, SpaceX, Relativity) run their own pipelines with different geographic and clearance constraints. Some hire heavily in Los Angeles, Denver, or Huntsville. Others recruit almost exclusively from specific university labs. There is no unified "frontier tech labor market" with a single clearing price or screening standard.

Finally, the sample is temporally thin. One role added in seven days. Four salaried roles total on the board. That snapshot could reflect a funding close, a contract award, or backfill after attrition. It does not, by itself, establish a trend line. The broader signal (domain fluency outweighing pedigree) is visible across multiple firms and quarters. Human Computer's current page illustrates it. The board is a map. The territory is the work itself — and the work is still being done in SCIFs, on flight lines, and in the latency budgets where models meet metal.

Working in frontier tech? Zero G Talent tracks the openings: see every open Human Computer role, browse frontier tech jobs, the companies hiring, and the people building the field.