500k AI Jobs Open Worldwide, Fathom.io Hires One RL Expert

A Single Role, A Wide Net

Fathom.io listed a Reinforcement Learning Expert role across Saudi Arabia, the United Kingdom, and Germany more than a month ago, a single opening at a 51–200 person company (LinkedIn's data shows) backed by Aramco's $500 million WAED Ventures fund (Wa'ed Ventures reported). The posting sits alongside a Rust Software Engineer position open in four countries, a Junior DevOps Engineer role in Saudi Arabia, and a Senior Data Scientist focused on computer vision, all marked "30+ days ago" on the WAED job board. That board also lists a Senior Scrum Master role (twice), a Junior Product Owner, and a generic "Open Position" in Saudi Arabia. Fathom's LinkedIn page shows two additional openings: Senior Talent Acquisition Specialist and Senior Organizational Development Specialist, both based at the Dhahran headquarters.

The reinforcement learning role stands out. Fathom describes its platform as an "Enterprise Intelligence Operating System", a composable, quantum-safe stack designed to run agentic AI systems at enterprise scale. The Composition Studio, powered by Data, Intelligence, and Application Workspaces, lets customers turn real-time intelligence into automated workflows. That architecture demands engineers who understand how learning agents explore, exploit, and generalize under constraints that academic benchmarks rarely capture: governance boundaries, data sovereignty rules, and the latency budgets of regulated industries like finance, healthcare, and smart-city infrastructure.

The company's footprint explains the geographic spread. Headquartered on King Saud Bin Abdul-Aziz Road in Dhahran's Al Qashlah district, Fathom maintains offices at 19 West 34th Street in New York, in Abu Dhabi, and in Oslo. CEO Ibrahim Al-Baloud represented the company at the Saudi House in Davos during February 2026, running a session on enterprise agility in the AI economy. In January 2026, the Saudi Minister of Communications and Information Technology visited Fathom's offices to review the platform. A memorandum of understanding with Think AI Labs, signed at LEAP 2026, signals active partnership building in the Kingdom's AI ecosystem.

A single reinforcement learning hire represents a meaningful capacity investment for a company of this size. The role's multi-country listing reflects a talent search that acknowledges where this expertise actually lives: reinforcement learning researchers with production experience cluster in specific labs and industrial teams; they don't distribute evenly across job boards. What the posting doesn't say is how Fathom screens for the intersection of RL theory, systems engineering, and the regulatory fluency that enterprise deployment demands.

Inside the Funnel

Glassdoor aggregates 52 interview questions and 49 reviews for Fathom, and a separate 3 questions and 3 reviews for Fathom Applications, suggesting the process is active and documented by candidates. A 1point3acres write-up describes a mid-level full-time SDE phone screen at Fathom. The hiring-manager stage calibrates level and team fit, shifting from pure technical execution to impact metrics, a conflict-or-recovery story, and a crisp "Why Fathom.io?" that references the specific business unit.

Timeline varies by track and geography. The research notes that country, business unit, and role family change stage order and tools, so the funnel contract is the live requisition itself: must-haves vs. nice-to-haves, role family, level signals, location and work-mode rules, and keywords that must be mirrored honestly on the résumé.

What Gets You Past the Screen

Fathom.io's interview data reveals a sharp split: Software Engineer (Data) candidates clear the process in one day on average, while Machine Learning Engineer applicants face a 26-day gauntlet, the slowest of any role at the company. That gap is the clearest signal of where the bar sits. The platform they're building — an Enterprise Intelligence Operating System™ described as "secure, quantum-safe, composable" with "intelligence built in by design, not bolted on" — demands engineers who can move beyond model training into governed, production-grade systems that unify data, decisions, and execution.

The technical stack implied by the mission is specific. "Quantum-safe" and "composable" point to cryptographic literacy and modular architecture experience. "Agentic AI systems built, governed, and evolved at scale" means Kubernetes operators, policy-as-code frameworks, and observability stacks that track decision provenance. A Glassdoor reviewer noted the company "does a good job of listening to new ideas and adopting better processes when applicable," suggesting the screen favors pragmatists who can argue for a better toolchain without derailing a release.

Experience at the intersection of enterprise software and frontier AI carries disproportionate signal. The recent memorandum of understanding with Saudi firm Think, signed at LEAP 2026, underscores a go-to-market motion tied to national digital transformation programs. Engineers who have sold into or built for sovereign-cloud buyers, defense-adjacent procurement, or Gulf-region megaprojects understand the stakeholder map that pure research hires do not.

Cultural fit reads less like a values poster and more like a survival trait. The organization spans Dhahran, New York, Abu Dhabi, and Oslo, four time zones, a 51–200 person headcount. The two Dhahran openings advertised in mid-2026 explicitly call for people "passionate about finding exceptional talent" and who "believe in the power of growth, capability building, and creating environments where people can perform at their best." The internal language mirrors what the engineering screen tests for: builders who treat process as a lever, not a constraint.

The 17-day average hiring cycle across 49 Glassdoor-reported interviews masks the variance. For the ML track, the 26-day average suggests multiple technical reviews and a system-design session grounded in Fathom's actual architecture. Candidates who arrive with a public track record — open-source contributions to orchestration frameworks, papers on governed AI deployment, or prior work at Aramco Digital, MOZN, Elm, or similar Gulf-region AI ventures — enter with context the interviewers don't have to teach. The screen filters for that context. It also filters for the patience to operate inside a company that describes its own product as "the digital foundation on which secure, adaptive, and agentic AI systems are built", a mandate that leaves no room for prototype-grade code.

The Market That Swallows Talent Whole

Fathom.io's single opening sits inside a market that has stopped behaving like a normal labor pool. LinkedIn's 2025 Jobs on the Rise list ranked Artificial Intelligence Engineer first in the United States. ManpowerGroup recorded a 16 percent jump in AI-related postings over a three-month span even as overall tech hiring fell 27 percent year-over-year. A separate recruiter survey logged a 59 percent surge in AI postings for 2024. The contradiction is structural: companies are cutting generalist roles while bidding up specialists.

The numbers are lopsided. IntuitionLabs counted roughly 500,000 open AI-related positions worldwide in 2025. Indeed's 2025 AI Hiring Index put active U.S. listings above 350,000, double the 2023 figure. Greenhouse measured a 23 percent year-over-year increase in San Francisco tech postings in the first quarter, averaging 2,190. Pragmatic Engineer found that 32 percent of all listed AI engineering jobs sit in the Bay Area, more than the next nine locations combined. Seattle ranked second with 1,472 postings as of January 2025.

Salaries have detached from the rest of software engineering. Greenhouse and Pave tracked a 53 percent climb in machine learning engineering pay over 15 months while general software engineer compensation crept up 4 percent. Median U.S. AI engineer pay now exceeds $138,000; top-tier offers routinely clear $300,000 with aggressive signing bonuses. Glassdoor's aggregate shows a typical base range of $86,000–$131,000, with Meta, Apple, and Microsoft reporting ranges up to $456,000. In London, principal machine learning engineers command £140,000–£300,000. Entry-level data scientist salaries jumped from $117,000 to $152,000 in a single year.

Big Tech absorbs 40 percent of all AI postings. Startups take 30 percent. Traditional industries — finance, healthcare, retail — claim the remaining 30 percent and represent the fastest-growing segment. Finance hiring for software engineers rose 91 percent in early 2025; industrial automation rose 73 percent. Seventy-two percent of Fortune 500 firms now run dedicated AI hiring initiatives. Microsoft recruited 24 researchers from Google DeepMind while simultaneously laying off roughly 9,000 non-AI staff. Meta's engineering headcount sits 19 percent above its January 2022 level. Google and Apple grew 16 percent and 13 percent respectively. Amazon and Microsoft have barely moved since early 2023.

The supply side cannot keep pace. Tech unemployment hovers near 2 percent. Thirty-five percent of companies cite high salary expectations as their top recruitment hurdle. Thirty-two percent of applicants overstate their AI expertise, according to Greenhouse and CloudZero surveys. Application volumes have swollen 239 percent since ChatGPT launched, flooding pipelines with low-intent spam. Hiring managers respond by lengthening interview loops, adding take-home projects, and demanding deeper verification, exactly the dynamic Fathom.io's screen reflects.

Remote listings are declining across tech except in AI engineering, where they are ticking up. More than half of open roles target senior level or above. The "AI Engineer" generalist title is fracturing into specialists: RAG engineers, LLM fine-tuning specialists, AI product managers who can translate transformers to executives. The EU AI Act and U.S. regulatory scrutiny have made AI ethics and compliance the fastest-growing niche.

First-party board data underscores the intensity. Anthropic added 54 roles in the past seven days with a salary band of $205,000–$550,000 (median $385,000). Databricks added 53 roles in the same window at $140,000–$317,000 (median $250,000). Both companies hire almost exclusively at staff and principal levels.

The talent shortage is not a cyclical dip. It is a structural mismatch between the capital required to train frontier models — Anthropic's Dario Amodei estimated $1 billion for 2024 training runs — and the narrow pool of engineers who can deploy them. Until model training costs drop an order of magnitude, the bidding war holds. Fathom.io's lone requisition is not an anomaly. It is the visible tip of a market that has priced most buyers out.

What Candidates Can Do Now

The hiring landscape Fathom.io operates in has shifted beneath every candidate's feet. Half of companies already use AI in their hiring process, and 68 percent will by the end of 2025, per a ResumeBuilder.com survey cited by CNBC. That means the screen you face at Fathom.io is not an outlier, it is the new baseline. The practical advantage goes to candidates who prepare for the format without trying to guess a proprietary scoring model. Your goal is not to "beat" an AI interviewer; it is to make your experience easy to understand.

Start by reading the invitation literally. "AI interview" is a broad label, and assuming every employer uses the same process is a mistake. Some companies run one-way video recordings scored by a model; others use conversational agents that probe follow-ups in real time; a growing number — Anthropic, Amazon, Deloitte's U.K. graduate program — explicitly forbid AI assistance during the assessment and ask candidates to acknowledge that constraint upfront. If Fathom.io's screen includes a live coding or prompt-writing component, treat it as a supervised exam. Interviewers have learned to spot eyes wandering to a second monitor, reflections of other apps in glasses, and answers that sound rehearsed or don't match the question asked. More than 50 percent of applicants in one virtual coding challenge cheated, according to the hiring manager who ran it, and the tools have only grown harder to detect. Getting caught eliminates you. The ethical boundary is simple: use tools to prepare, organize your thinking, and practice; follow the employer's instructions during the actual assessment.

Preparation should center on three concrete artifacts. First, prepare two to three specific stories that demonstrate fluency, not credential matching. Hiring managers in 2026 ask you to walk through a real problem you solved with AI, explain how you would approach a new one, and show you can reason about where AI helps and where it does not. Technical depth varies wildly by role; honesty about your level is an asset. For career changers, your domain expertise is part of the value proposition, not a liability. Second, practice a live prompt-writing exercise: time yourself writing a prompt for a task the role would actually care about. Third, build your "domain × AI" framing: why does your background make you better at this than someone without it? The candidates who do best are the ones who have actually been building: running real experiments, documenting what worked and what didn't, iterating.

If a disability, technology constraint, or another barrier could affect your ability to complete the interview, ask the employer promptly about an accommodation or an alternative process. The EEOC and DOJ have made clear that employers' use of software, algorithms, and AI in employment decisions must comply with civil-rights protections including the ADA. The EEOC's technical assistance document and its "Tips for Job Applicants and Employees" summary are public resources worth reviewing before you engage.

Finally, assume the process will evolve. Google's CEO has suggested a return to in-person interviews; startups like Henry Kirk's are considering the same move. A single role at Fathom.io may attract hundreds of applications, and the screen you clear today could look different in six months. The preparation that carries forward — whether a bot or a human recruiter joins next — is confirming the format, bringing specific evidence, rehearsing your delivery, and leaving enough room for your genuine voice.

Outside the Frame

This article examines a single engineering opening at Fathom.io and the screening criteria attached to it. That focus deliberately excludes several adjacent topics that often appear in hiring coverage but fall outside the reported scope.

The piece does not analyze Fathom's broader compensation architecture. The research shows a company that reached $30 million ARR in 2025 on roughly $30 million total capital raised, revenue roughly equaling every dollar ever invested, and a CEO who said he runs the company with "probably never more than a million in the bank." Those financial dynamics shape what the company can pay, but the article does not break down salary bands, equity grants, or how the $25-per-seat average revenue translates into engineering offers. Zero G Talent's board data shows Databricks roles in a $140k–$317k band and Anthropic roles at $205k–$550k; Fathom's figures are not in that dataset and are not speculated on here.

It does not compare Fathom's hiring bar to competitors. The research lists Gong ($500M revenue, 2,248 employees), Otter.ai ($100M revenue, 200 employees), Fireflies.ai ($10.9M revenue, 100 employees), Avoma ($15M revenue, 62 employees), Grain ($10.3M revenue, 72 employees), and others. Each runs its own interview loops, calibration practices, and role definitions. This story treats Fathom's screen as a standalone signal, not a benchmark for the category.

It does not reconstruct Fathom's technical architecture. White told TechCrunch the company relies on its own fine-tuned models and that "working with models is very different from typical engineering projects", output is "not a feature, it's actually spec" with a "50% time right now" failure rate. Those details explain why the role exists, but the article does not dive into model selection, inference infrastructure, the undocumented APIs used for capture across Zoom, Google Meet, and Teams, or the new botless capture modes, MCP server, or local file-system export White previewed in May 2026.

It does not evaluate the product roadmap beyond what the hiring need implies. The research describes Ask Fathom, a query layer over organizational meetings, plus AI scorecards, automated follow-up emails, CRM agent actions, and first-party Claude/ChatGPT integrations. Those features justify hiring, but the story does not assess their feasibility, timeline, or market fit.

It does not cover Fathom's fundraising history in detail. Sources conflict: Tracxn shows four rounds including a Series B in November 2022 and another in May 2026 with amounts redacted; GetLatka and StartupIntros describe three rounds totaling $30M with a $17M Series A at a $73M post-money in June 2024; TechCrunch reported a $17M Series A in September 2024 led by Telescope Partners with $2M from Wefunder crowdfunding. The discrepancies are noted in the research but not adjudicated here, they do not change the screening criteria for one engineering role.

It does not address the "8 customers" figure that appears in some profiles. The GetLatka tape clarifies that number is the average seat count per account (8–10), mis-slotted as a customer count; at $30M ARR and ~$200/month per account, the real account count is orders of magnitude higher. The article does not correct every data error in circulation.

It does not profile Richard White's career beyond the facts that bear on hiring. His twelve years at UserVoice (peak ~$10M revenue on ~$9M raised), the Kiko/Y Combinator origin, and his "attack metrics in order of risk" sequencing, retention, activation, acquisition, referral, monetization last, inform how he builds teams, but the story is not a founder biography.

It does not explore the freemium-to-teams monetization model, the decision to keep individual note-taking free, the $100K ARR first month after a year of zero revenue, the 800-person manual onboarding experiment, the Zoom App Marketplace launch forcing function, or the sales-team hiring a year before monetization. Each shaped the company that now posts one engineering role, but they are context, not subject.

It does not cover visa sponsorship, relocation, remote-work policy, onboarding structure, retention rates, promotion ladders, or diversity metrics. The research does not contain those specifics for Fathom, and the article does not extrapolate.

It does not assess the broader AI talent market beyond the observation that a single role at a $30M ARR, 100-person company draws intense competition. The Databricks and Anthropic board snapshots illustrate the pressure, but the story does not attempt a labor-market analysis.

In short: this article reports what the screen for one role at Fathom.io looks like, why it looks that way, and what candidates can do about it. Everything else, compensation philosophy, competitor practices, technical deep-dives, fundraising forensics, founder narrative, product strategy, market sizing, and policy details, is outside the frame. The role has been open 30-plus days. The screen is still running. The next candidate who clears it will join a team building the digital foundation for agentic AI at enterprise scale, one reinforcement learning specialist at a time.

Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.