90% of U.S. Firms Use AI Screening — Human Computer Tightens

Six Roles, Three Hubs, One Bet

Human Computer posted nine new openings across AI, robotics, and systems engineering in a single week, expanding its careers board across three continents and setting a published salary band with a $150,000 median, per its careers page and Zero G Talent's board data. The move signals a shift from maintenance-mode backfilling to a deliberate build-out, and it triggered an immediate surge in applicant volume. Six of the nine roles appear on the Zero G Talent board; the table below captures them.

*Posted within the board's published band; no explicit range listed.

Zero G Talent's figures put the Senior Security Researcher range at $135,000–$165,000.

Geography is deliberate. Tel Aviv has long been the company's engineering center of gravity; adding design and people ops there signals product velocity is about to increase, not just headcount. London's customer success hire points to enterprise motion: someone who can translate technical depth for buyers who don't speak kernel. The U.S. security roles map to a threat landscape that has shifted from perimeter defense to model-level integrity. Each location carries a different regulatory and talent-market reality; Human Computer is hiring into all three simultaneously.

The roles also reveal a departmental architecture that doesn't map to traditional org charts. Security research sits beside backend engineering. UX/UI reports into product but shares a hub with people ops. Customer success engineering — a hybrid of solutions architecture and post-sales engineering — lives in London while the product it supports is built in Tel Aviv. This isn't a company adding headcount to existing teams. It is standing up a cross-functional spine that can ship, secure, and support a product category without a standard playbook.

Only two of the six new roles carry explicit salary ranges; the rest sit inside the board's published band. That pattern means either the band is still being calibrated or the company expects to negotiate individually for profiles that don't fit standard ladders. In frontier tech, it means the latter.

The timing coincides with a broader squeeze. Deloitte's 2023 tech talent survey found only 13% of employers believe they can hire and retain the talent they need most, while 72% of U.S. tech employees say they're considering a move. Human Computer's announcement lands inside that squeeze, not as a reaction but as a bet that the right nine people, placed in the right three hubs, can outproduce a far larger team hired to spec.

Inside the First Filter

Ninety percent of U.S. employers now use AI screening tools to sort and rank job seekers. Stanford HAI researchers tracked 3.4 million people submitting 4 million applications across 1,700 postings and found most rely on the same few third-party vendors, creating what they call algorithmic monocultures: a single vendor's model can gatekeep access to dozens of companies at once. The same study found the system had already filtered 10% of four-time applicants into universal rejection.

The first screen is rarely human. Self-recorded video interviews have become standard. CNBC reported that candidates face a camera, a timer, and no interlocutor. The system scores keywords, facial micro-expressions, speech cadence, and eye contact. Nature research on AI-enabled interviews shows applicants feel scrutinized at a granular level but lack the knowledge to effectively respond or optimize their performance. Perceived procedural justice drops; organizational attractiveness follows.

What the models actually reward remains opaque. Stanford researchers found that pooling a vendor's recommendations across all positions masks adverse impact; however, evaluating each position separately, as adverse-impact law requires, exposes discrimination in many roles. Twenty-six percent of Black applicants and 15% of Asian applicants applied to positions where the AI system discriminated against their racial group. If the AI had recommended those candidates at the same rate as the most-favored group, 40,000 more applications would have advanced. The bias stems from limited training data and biased designer choices, creating a "bias in, bias out" loop that modern algorithms can amplify into what researchers term agentic discrimination: seemingly neutral outputs that disproportionately harm protected classes.

Candidates who clear the screen share a narrow playbook. Career advisors recommend the STAR method — situation, task, action, result — to structure every anecdotal response. Look directly into the camera, not the screen. Prepare outfit and background. Get to the point. A widely circulated coaching framework summarizes it as "always be chatting": keep talking, avoid language-specific magic, and if you freeze, ask a question rather than sit silent. These are performance tactics for an algorithmic audience, not conversation skills for a human evaluator.

Human Computer's tightened criteria arrive as applicant volume surges. Each role demands hybrid fluency: systems engineers who read research papers, security researchers who write production code, designers who understand model limitations. The screening must separate performers from performers-of-performance. Candidates now tailor resumes to inferred keyword weights, upskill in the specific domains the job descriptions signal, and rehearse STAR stories for a camera that does not blink.

The Market Responds

China's core AI industry exceeded 1.2 trillion yuan in 2025 (roughly $173 billion), spanning more than 6,200 enterprises, according to official data released in March 2026. That scale frames the labor market Human Computer now competes in.

Recruiters at frontier-tech firms report a measurable shift in candidate behavior since the announcement. Applicants are foregrounding hybrid skill sets that mirror Human Computer's posted requirements: large-language-model integration paired with real-time control systems, simulation-to-reality transfer for robotics, and safety-critical software practices borrowed from aerospace. The board's listings — Senior Security Researcher, Threat Research Analyst, Backend Engineer in Tel Aviv, Customer Success Engineer in London — read like a map of the competencies now in shortest supply.

Government policy is amplifying the trend. The State Council's August 2025 "AI+" guidelines named embodied intelligence as a priority growth driver for the 15th Five-Year Plan period (2026–2030), targeting an AI-related industry scale above 10 trillion yuan. Beijing followed with a dedicated action plan to build a global brain-computer-interface innovation hub. Those directives translate directly into curriculum changes: universities are adding BCI-AI integration modules, and bootcamps are compressing autonomous-agent frameworks like LangChain and Auto-GPT into weekend sprints.

The skills gap shows up in training infrastructure, not just resumes. West Virginia University's GOLab now runs power-grid crisis simulations where 60 engineering students operate as control-room operators while sensors track skin conductance and eye movements, a direct analogue to the human-machine teaming Human Computer's systems roles demand. "Operators typically work 12-hour shifts," said researcher Anurag Srivastava. "If they receive too much information, they can't process it. Too little, and they make the wrong calls." That calibration problem — information density versus cognitive load — is exactly what the new Backend Engineer and Security Researcher hires will face on day one.

Candidates are responding. GitHub data shows forks of autonomous-agent repositories up sharply in the last quarter; Hugging Face model cards increasingly tag "real-time" and "embodied" alongside standard LLM benchmarks. The shift isn't cosmetic. A Threat Research Analyst who once focused on static model extraction now needs red-team experience against agents that can chain tools, browse, and persist across sessions. A UX/UI Designer in Tel Aviv must prototype interfaces for operators who supervise fleets, not single robots.

The ripple extends beyond direct applicants. Competing firms, especially those in defense-adjacent robotics and critical-infrastructure automation, are raising their own bars to retain talent Human Computer might poach. Salary transparency on the board makes that dynamic visible: when a Customer Success Engineer role in London lists at the top of the band, it resets expectations for similar posts at peer companies.

A tightening feedback loop has formed. Policy targets create funding; funding creates roles; roles demand hybrid fluency; candidates invest in that fluency; the talent pool deepens but also narrows around a specific profile. Human Computer's screen sits at the choke point.

Benchmark: Where Human Computer Stands

The frontier-tech hiring market is crowded. As of February 2026, BuiltIn listings show NVIDIA advertising 1,795 openings, SharkNinja 245, General Dynamics Mission Systems 85, Nuro 78, Apptronik 64, Agility Robotics 52, Boston Dynamics 42, Ouster 27, MORSE Corp 29, Carbon Robotics 15, and iRobot 10, alongside "all jobs" listings for Anduril, Tesla, Amazon Robotics, Starship Technologies, and SVT Robotics. Candidates interviewing at Human Computer are almost certainly interviewing at three or four of these companies in the same cycle.

Salary bands offer the clearest public benchmark. Human Computer's published salary band puts the median salaried offer at $150,000 with a $95,000–$162,000 range, per Zero G Talent's board data. Those figures sit in the middle of the defense-and-autonomy tier: above pure-play robotics startups that often top out near $140,000 for senior ICs, but below the $180,000–$250,000 bands NVIDIA and Tesla routinely post for comparable AI infrastructure and systems engineering roles. The gap reflects a structural difference: Human Computer's roles skew toward security research and backend engineering, while peers hiring for large-model training infrastructure or vehicle autonomy command a premium.

Screening mechanics are harder to compare because few firms publish their rubrics. Anduril is the exception that proves the opacity: its AI Grand Prix drone-racing contest, announced in January 2026, replaces the first technical screen with a live autonomy challenge. Teams write the software that flies a Neros Technologies drone through a closed course; the highest scorers bypass the standard recruiting cycle entirely. The $500,000 prize pool and direct-hire pathway make the contest a screening instrument as much as a marketing event. Luckey framed it as a philosophical necessity: "The whole point... is this pitch that autonomy has finally advanced to where you don't have to have a person micromanaging each drone." Human Computer has not adopted a public challenge model; its tightened screen appears to rely on traditional technical assessments and portfolio review, putting the burden on candidates to self-select into the right preparation.

The rest of the peer set, including Boston Dynamics, Agility, Apptronik, Nuro, and Amazon Robotics, recruit through conventional loops: phone screen, take-home or live coding, system design, onsite. Public signals suggest the bar is rising uniformly. NVIDIA's interview packets now routinely include CUDA optimization and distributed training questions that were PhD-qualifier material five years ago. Agility and Apptronik test whole-body control and sim-to-real transfer on hardware-in-the-loop rigs. Human Computer's emphasis on security research and backend scaling aligns more with Anduril's and MORSE Corp's defense-adjacent stacks than with the mobility-focused peers, which means candidates carrying ROS 2, LCM, and real-time Linux experience have a portable advantage across that subset.

What the market lacks is a shared vocabulary for "hybrid" competence: the blend of ML model deployment, embedded systems, and safety-critical software that Human Computer's nine roles collectively demand. Peers define it differently: Anduril calls it "autonomy engineering," Nuro "robotics software," Amazon Robotics "manipulation and perception." The terminology drift forces candidates to translate their experience for each application. Human Computer's screening tightening, whatever its exact thresholds, is effectively a bet that a narrower, well-defined rubric will outperform the industry's fuzzy defaults. Whether that bet pays off in hire quality or just narrows the funnel further won't be visible until the first cohort ramps.

Outlook: The Cohesion Problem

Human Computer's current hiring wave — nine openings spanning AI, robotics, and systems engineering — signals a product roadmap hardening around three vectors: autonomous systems security, distributed compute infrastructure, and human-in-the-loop interfaces. The board's salary band positions the company above early-stage startup norms but below the hyperscaler ceiling, a deliberate calibration that attracts senior individual contributors who have already survived one acquisition or pivot.

The composition of the open roles tells its own story. A Threat Research Analyst and a Senior Security Researcher suggest the next product cycle will harden against the corporate espionage trend Gartner flagged for 2026: insider threats and AI-powered espionage moving "from the pages of fiction to payrolls." Meanwhile, the Customer Success Engineer in London and the UX/UI Designer in Tel Aviv indicate a push toward deployed systems that require ongoing field support, not just lab demos. That shift from research to operations is where most frontier-tech companies lose momentum. Gartner's 2026 data shows only one in 50 AI initiatives delivers transformative value, and fewer than 1% of those gains trace to actual productivity improvements. The rest stall in what Gartner calls "workslop": low-quality output that employees spend hours fixing.

Retention will hinge on whether Human Computer can avoid the culture dissonance Gartner identifies as a 2026 risk: stated values that don't match day-to-day realities. The biotech sector offers a cautionary contrast. In August 2026 alone, BioNTech shuttered a Berlin subsidiary (140 roles), Pfizer added $2.5 billion in cuts via "technology and simplification efforts," and Novartis trimmed its East Hanover site for the fourth time (322 more roles). Those cuts reflect clinical and commercial failures, not hiring miscalculations, yet they flood the market with specialized talent that Human Computer's security and backend roles could absorb if the compensation narrative holds.

The legislative environment adds another constraint. California's 2025 No Robo Bosses Act and Massachusetts' FAIR Act now require transparency and human-in-the-loop guardrails for algorithmic management. Colorado's 2024 AI Act set the precedent. A federal Right to Override Act would let healthcare workers override care-directing algorithms. Human Computer's People Operations Specialist hire (maternity leave coverage in Tel Aviv) suggests the company is building HR infrastructure ahead of regulatory pressure, not reacting to it.

Gartner's 2026 research finds that teams redesigning workflows with AI are twice as likely to exceed revenue goals, but only when "process pros, not tech prodigies" lead the redesign. Human Computer's blend of security, backend, and customer-facing roles suggests it understands that equation. The 15-person cohort now forming will test it. The board's median $150,000 buys competence; it doesn't buy cohesion. That comes from shipping, and the careers page that moved this fast in one week will be the first place to show whether it holds.

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.