CX2’s $225K AI Jobs Require RF Hardware, Not Models

What CX2's Five Roles Are Actually Screening For

The job postings don't ask for machine-learning engineers. They ask for people who can make RF hardware survive a jammer.

CX2, a Series A defense-technology company backed by Andreessen Horowitz and led by founders with track records at Meta, SpaceX, Epirus, and the U.S. Department of Defense, lists five salaried openings on the Zero G Talent board as of October 2026. All sit in El Segundo, California. The roles, Lead RF Hardware Engineer, Software Engineer - Software Infrastructure (Cybersecurity), Electrical Engineer, Signal Processing Engineer, and EW Hardware Engineer, share a single mission: detect, disrupt, and defend the electromagnetic spectrum across land, air, sea, and space. CX2's screen rewards candidates who can translate commercial machine-learning experience into contested electromagnetic warfare — not model accuracy, but latency, size-weight-power constraints, and graceful degradation under jamming.

The Electrical Engineer posting makes the screen explicit. Required qualifications demand a bachelor's or master's in electrical engineering and three-plus years of electronics design, defense or aerospace preferred. The hard-skills checklist runs from analog and digital circuit design through PCB layout in Altium or KiCad, embedded microcontrollers, DSP systems, power and battery management, to lab equipment: oscilloscopes, spectrum analyzers, signal generators. The posting favors candidates with counter-UAS, radar, or electronic-warfare experience; knowledge of RF jamming, spectrum monitoring, or direction finding; FPGA familiarity; and military standards such as MIL-STD or DO-160. The role also demands an existing U.S. security clearance or the ability to obtain one, U.S. citizenship or green-card status per export regulations, and willingness to work extended hours and weekends for "mission critical deadlines."

The Lead RF Hardware Engineer and EW Hardware Engineer listings reveal the same priority: hardware that operates inside a denied spectrum. The Signal Processing Engineer role targets the algorithmic layer, turning raw RF into targeting data fast enough for autonomous drones. The Software Infrastructure (Cybersecurity) opening is the outlier, but its placement beside the RF roles shows CX2 treats the software stack as a hardening problem, not a modeling playground. A separate Head of Manufacturing search on Jobera reinforces the hardware-first cadence: "lead and scale our manufacturing organization… highly hands-on leadership role for someone who understands how to manufacture complex hardware in a fast-paced environment."

None of the five descriptions lead with PyTorch, TensorFlow, or large-language-model experience. They lead with spectrum analyzers, MIL-STD qualification, and the ability to integrate sensors — radar, EO/IR, acoustic, RF — into platforms that integrate with existing command-and-control tools. The screen filters for engineers who have already worked inside the constraints of tactical networks and export-controlled programs. Commercial AI credentials appear only if they sit on top of that foundation.

The Translation Gap: Commercial AI vs. Contested Spectrum

CX2's five open roles tell the story before a recruiter reads a single bullet. Four are explicitly RF, signal processing, or electronic warfare hardware positions. The fifth sits in a cybersecurity context, not a generic ML platform one. This is not a company hiring "AI engineers" in the commercial sense. It is a company building hardware-defined, spectrum-dominance systems where AI — if it appears — runs on DSP chains inside adversarial electromagnetic environments.

The vocabulary mismatch is structural. A commercial AI résumé in 2026 leads with PyTorch, TensorFlow, LLM fine-tuning, RAG pipelines, MLOps on Kubernetes, and cloud inference optimization. CX2's screen rewards SIGINT collection management, Indications and Warning (I&W) and Attack Sensing and Warning (AS&W) workflows, network target development, and digital signal processing on RF hardware. The Principal Technical Director, AI-Enabled Spectrum Dominance role advertised across the industry, driving Voyager's Agentic Computational Engineering (ACE) platform into RF hardware and DSP development, makes the distinction explicit: the AI is a means to spectrum dominance, not the product itself. The same language appears in Senior Technical Director, Spectrum Dominance postings seeking leaders for "advanced RF systems, software, and algorithms portfolio" in electromagnetic warfare and sensing.

Veterans who have spent a decade and a half as Electronic Warfare technicians managing intelligence operations, supervising hundreds of personnel, and holding active TS/SCI clearances possess the raw material. But the translation gap has three layers. First, military titles, "Electronic Warfare Technician," "Program Manager for Electronic Warfare team," do not map to civilian search strings. Best Military Resume identifies about 30 military titles a civilian would never search. Second, duty lines written for military evaluation boards (FITREPs, NCOERs, OPRs) answer "what did you do" but rarely answer "so what" in commercial terms, dozens of such lines per resume, by their count. Third, service jargon, "SIGINT," "I&W," "AS&W," "network target development," forces a hiring manager to guess the equivalent civilian capability. About 25 such terms per resume create friction a lean recruiting desk in 2026 will not absorb.

The defense industry's own postings confirm what CX2's roles imply. Boeing Defense seeks Electromagnetic Effects Design and Analysis Engineers for E-7A and KC-46A programs, E3 (Electromagnetic Environmental Effects) analysis, not model training. The U.S. Air Force's Cyber and Electromagnetic Activities specialty requires frequency assignment, interference resolution, and spectrum certification for radiating equipment. Major contractors, Northrop Grumman, Raytheon Technologies, L3Harris Technologies, concentrate R&D in California, Florida, and Texas specifically for RF spectrum dominance platforms. Industry sources describe the electromagnetic spectrum as "one of the most contested operational environments in modern defense and critical infrastructure protection."

For a candidate coming from commercial AI, the reframing required is not cosmetic. You must demonstrate you have touched RF hardware — SDRs, phased arrays, signal generators, spectrum analyzers — and understand the physics of a denied band: jamming, LPI/LPD waveforms, cognitive radio adaptation under adversarial interference. You must show you can move an algorithm from a Jupyter notebook onto an FPGA or a DSP core with deterministic latency, size-weight-power constraints, and no cloud fallback. You must articulate how your ML work survives when the spectrum is denied, degraded, or spoofed. And you must hold or be eligible for the clearance that lets you walk into the SCIF where the real data lives.

The screen rewards evidence over keywords. Bullets that quantify results in denied-spectrum terms, referencing specific hardware targets, SNR under jamming, and clearance status, speak the language of the Signal Processing Engineer and EW Hardware Engineer roles. The principle is the same one that turns "managed a fleet" into "managed a fleet valued above $40M" and "improved readiness" into "moved deadlined equipment ready rate from 71% to 89%."

Candidates who clear this screen have done the translation work: they map commercial ML experience onto the specific RF/DSP/EW problems the job descriptions name, strip service-only tool names unless the posting calls for them, and lead with the one detail a civilian hiring manager cannot pull from a generic summary, an active clearance, a rare RF specialty, or a demonstrated hardware-to-software integration in a denied-spectrum environment. The gap is real. The candidates who cross it stop writing AI résumés and start writing spectrum-dominance résumés that happen to use AI.

Clearance, Citizenship, and the El Segundo Filter

El Segundo is not a suggestion. It is the constraint. CX2's five open roles are all seated in this single Los Angeles County city. Founder Nathan Mintz built the company here for a reason: El Segundo sits at the heart of the Southern California aerospace corridor, anchored by Space Systems Command at Los Angeles Air Force Base, which develops and acquires the nation's military satellite constellations. The area hosts one of the densest concentrations of satellite and aerospace primes in the country, plus a fast-growing defense-startup presence.

The clearance picture is more nuanced than the job boards suggest. Public postings on MissionRobo show "Clearance: none" for every CX2 role. But "none" on a public requisition does not mean "unrestricted." Defense electronic warfare work, especially on autonomous drones delivering distributed EW capabilities for spectrum dominance across those domains, almost always requires U.S. person status under ITAR and EAR. The company's own careers page states its mission: "empower the United States and its allies to dominate the electromagnetic spectrum." That language carries citizenship implications regardless of whether a specific requisition lists an active clearance. Candidates who are not U.S. persons (permanent residents included, depending on the program) will hit a hard wall at the offer stage or during onboarding when program access is adjudicated.

Labor market data confirms this. Cleared defense engineering and program management vacancy rates in El Segundo remained above 20% throughout the recent layoff period. The clearance barrier creates separate labor markets: commercial aerospace slack does not transfer to classified defense shortages. A software engineer laid off from a commercial satellite constellation operator faces the same citizenship and clearance barriers. The geographic concentration compounds the filter. El Segundo's talent pool is deep but narrow: specialized RF, signal processing, and EW engineers who are also U.S. persons and willing to work on classified programs are a finite set. Recent advances in AI-driven electronic warfare, spectrum congestion, autonomous systems, and next-gen wireless comms force contractors and infrastructure operators to compete for this same talent slice.

Deloitte reports that competition for AI-trained employees will intensify as aerospace and defense businesses increasingly incorporate autonomous systems and AI into their products and operations. CX2 is hiring into that pressure. The five roles span the stack from RF hardware through signal processing to cybersecurity infrastructure, each a discipline where commercial experience exists but denied-environment credibility is scarce. The "El Segundo filter" is not just geography; it is the intersection of citizenship, clearance eligibility, domain-specific RF/EW experience, and the ability to operate on defense acquisition timelines. The roles are on-site. The mission is classified-adjacent. The talent pool is the cleared, U.S.-person, Southern California RF/AI intersection, and it is already oversubscribed.

How Candidates Are Rewriting Their Profiles to Pass

The job descriptions make the priority explicit. The Lead RF Hardware Engineer role asks for experience with LPI/LPD waveforms and anti-jamming techniques. The Signal Processing Engineer posting emphasizes real-time DSP on FPGA or embedded GPU targets. The EW Hardware Engineer description calls out electronic attack and protection systems.

Open-architecture fluency has become a standalone credential. The Army's SBIR award to HERE for an AI-agnostic integration layer, browser-based, REST-API-driven, designed to sit atop legacy C2 systems without replacing them, mirrors the environment CX2 operates in. Candidates surface work with TacticalMesh or similar decentralized mesh stacks: multi-hop routing between agents, offline command buffering with sync-on-reconnect, Prometheus metrics export for observability in denied environments. Experience deploying on Raspberry Pi, Jetson, or generic Linux edge servers reads as direct preparation for CX2's commodity-hardware requirement. Contributions to the TacticalMesh repository (provided they avoid export-controlled or classified code) demonstrate dual-use, open-system thinking.

Rapid adaptation, the other half of CX2's mission, shows up in development-cycle descriptions. The vocabulary shifts from DevOps to DevSecOps with an emphasis on the "Sec": security accreditation, Authority to Operate pathways, handling of controlled unclassified information. Candidates who have navigated a DIACAP or RMF process, even on a small program, flag it prominently.

Interoperability experience carries weight. The Spectrum Dominance Division at VTNSI, three dozen faculty researchers across electrical engineering, computer engineering, computer science, and applied mathematics, signals the multidisciplinary depth CX2 expects. Applicants highlight cross-domain projects: radar and SIGINT fusion, SAR/ISAR image classification feeding electronic attack decision logic, coordinate transformation between disparate sensor reference frames.

The El Segundo filter reinforces the rewrite. Physical presence in the Los Angeles aerospace corridor (or a credible relocation plan) appears alongside clearance status. Candidates reference local ecosystem touchpoints: SpaceX, Northrop Grumman, Raytheon, the Aerospace Corporation, UCLA's wireless research groups. The message is logistical as much as technical: readiness to start, familiarity with the commute, awareness of the security office.

None of this guarantees an interview. The screen remains opaque by design. But the consistent pattern: every commercial credential gets a defense translation layer. The model architecture matters less than the deployment constraints it survived. The dataset size matters less than the electromagnetic environment it was collected in. The cloud platform matters less than the radio hardware it eventually drove. CX2 is hiring for spectrum dominance. The candidates who pass are the ones who prove they have already operated in that battlespace, or have built the closest civilian approximation they can document.

What This Hiring Pattern Does—and Does Not—Mean

CX2's five open roles sit at a precise intersection. They demand fluency in open-architecture command-and-control tools, denied electromagnetic environments, and defense acquisition timelines. That combination filters for a profile the broader market is still learning to produce.

The pattern signals three shifts. First, spectrum dominance has moved from a niche specialty to a hiring priority at venture-backed defense-AI firms. Second, the screen rewards translating commercial ML experience into electronic warfare, the same constraints as before. Third, the El Segundo base creates a geographic filter that compounds the clearance filter. Southern California aerospace talent pools are deep, but the overlap between cleared RF engineers and production-grade ML practitioners remains thin.

What the pattern does not mean: a general AI hiring boom in defense. The sector attracted $49.1 billion in 2025 venture funding, nearly double 2024, concentrated in a handful of companies. Anduril's $2.5 billion Series G and Saronic's $600 million round distort averages. Most defense-AI startups are still under-50-person teams hiring one or two specialists at a time. CX2's five roles represent a meaningful cluster, not a wave.

It also does not mean salary parity with commercial AI. Zero G Talent's data shows CX2's bands run $131k–$225k. That sits below the $200k–$300k+ packages large language model labs offer, but above traditional defense prime contractor ranges for equivalent experience. The gap reflects a specific trade-off: mission access and hardware-in-the-loop velocity versus pure cash compensation.

The hiring pattern does not reveal contract financials or program-of-record status. CX2's public job descriptions reference open-architecture C2 and rapid adaptation, language that maps to middle-tier acquisition and software acquisition pathways, but no posting discloses contract values, periods of performance, or transition plans. Inferring revenue trajectory from role count would be speculative.

Finally, the pattern does not generalize to other vendors' processes. Shield AI runs a five-stage loop heavy on C++ and domain-specific technical rounds. Anduril emphasizes autonomy stack integration. FTI Defense leans on modeling, simulation, and wargaming fluency. Each company's screen reflects its product architecture and customer set. CX2's emphasis on RF hardware, signal processing, and electronic warfare hardware maps to a specific mission set — spectrum dominance — that other defense-AI firms approach differently or not at all.

What the pattern confirms: the knowledge half-life in defense AI has compressed to months. The screen selects for translation ability, not just technical depth.

The broader implication: defense-AI recruiting is bifurcating. One track hires cleared engineers with legacy domain knowledge and teaches them modern ML. The other track hires commercial ML practitioners and teaches them electromagnetic warfare. CX2's roles sit in the narrow overlap. That overlap will widen as more venture-backed firms field spectrum-capable systems, but as of late 2026 it remains the bottleneck.

Working in frontier tech? Zero G Talent tracks the openings: see every open CX2 role, browse frontier tech jobs, the companies hiring, and the people building the field.