How Work Gets Done at Protex AI
Protex AI deploys model updates daily across construction sites, warehouses, and manufacturing floors. The company operates with a mandatory two-week shadow mode and a 99-percent precision threshold before any alerting rule goes live. A remote-first team of roughly 60 people spans Dublin, Budapest, Boston, and a Limerick hub listed on the careers page, with seven live openings on the Zero G Talent board from Senior DevOps Engineer in Budapest to Engineering Manager, Infrastructure in Dublin, all tagged remote-eligible. Founders Dan Hobbs (CEO) and Ciarán O'Mara (CTO) set product direction from Dublin. The technical architecture lets models adapt to a new site through configuration instead of retraining. Engineering leads enforce the precision threshold and shadow-mode validation. When customers flag false positives, those examples go straight into the refinement pipeline; improvements typically reach production within hours to days.
Models come pre-trained on industrial environments. During a dedicated calibration period the system tunes to a site's lighting, camera angles, and layout. Updates then deploy daily automatically. The platform tracks precision, recall, false-positive rate, and a relevancy metric the team says matters more than raw accuracy. Real-time alerting rules suppress repeated detections within a defined window, filter below confidence thresholds, and route to specific people rather than broadcasting. The goal, per product documentation: every alert that reaches a human means something.
Decision-making mirrors the product loop: short cycles, direct feedback, individual ownership. The careers page lists "Own it. Act with Urgency" as the first value — "we think like owners (not renters) and both empower and encourage autonomy." "Adapt, Overcome, Thrive" adds comfort with being uncomfortable and the belief that constraints drive creativity. Technical Solutions Engineers in Dublin or Limerick work with customers who adjust zones, thresholds, dwell times, and object sensitivity through visual interfaces. Customer Success Managers in Boston collaborate with engineering on site-specific detection gaps. Enterprise Account Executives in Dublin and Boston carry quota and feed frontline operational reality back into the roadmap.
Privacy and compliance are baked into the architecture. Event detection, blurring, and encryption happen locally on edge devices; raw CCTV footage never leaves the site. Only short encrypted clips and metadata reach the cloud. Four anonymization modes — facial blur, full-body blur, silhouette ghosting, and videoless mode, are configurable per customer. Access is role-based with MFA and logged. The stack holds ISO/IEC 27001:2022 (independently audited annually), GDPR, UK DPA, CCPA compliance, and alignment with the EU AI Act (classified as non-high-risk). TLS 1.2+ in transit, AES-256 at rest.
A startup moving at daily-deploy velocity sells into factories that cannot shut down for rewiring. Safety leaders need 80-percent near-miss reductions in 54 days. The values that drive this pace are explicit, and they show up in the interview room.
Values That Drive Decisions
Protex AI publishes its values directly on its careers page: keep growing, deliver excellence, treat feedback as a gift, and push beyond "good enough" through curiosity and innovation. The company also states that diverse perspectives drive innovation. These appear alongside the expectation that individuals hold themselves and each other to the highest standards. Protex AI keeps the list concise enough to remember, which matters in a remote-first team where cultural shorthand replaces hallway osmosis.
Founder Dan Hobbs described the same themes on the SaaStock podcast. He traced "keep growing" to his first startup, Better Examinations, where he bootstrapped for years and once survived on store-brand Rice Krispies for three weeks. That experience taught him that speed of learning compounds faster than any single technical advantage. The same interview revealed how "deliver excellence" translates to enterprise sales: the team prepares for Fortune 500 procurement reviews the way a safety-critical engineering team prepares for a regulator — every claim backed, every integration path mapped.
Co-founder Ciarán O'Mara has spoken less publicly. The engineering culture treats computer vision models as safety infrastructure. The value "feedback is a gift" shows up in model-performance reviews where false-positive rates are discussed openly across product, engineering, and customer success. Hobbs said the AI tools the team adopts internally are evaluated on whether they shorten the loop between hypothesis and evidence, not on vendor hype.
The remote-first structure amplifies the need for explicit values. With hubs in Dublin, Budapest, Boston, and a distributed U.S. presence, the company cannot rely on physical proximity to align priorities. Hobbs described the "boots on the ground" principle: senior hires in the U.S. came after the team learned that remote-only enterprise selling failed to build the trust required for multi-million-dollar contracts. That decision — hiring locally while keeping the engineering core distributed, illustrates how the values resolve tension between speed and control. The company still optimizes for shipping velocity, now defined as time to value for a safety team rather than commits per sprint.
A candidate evaluating Protex AI should read the values page, then watch the founder interviews. The overlap is high. The gaps — where a value feels generic until a specific story pins it down, are where the real culture lives.
What the Hiring Bar Reveals
The interview questions Protex AI publishes tell you more about the bar than any values page. Five questions, all grounded in shipping software that runs in production environments where safety incidents carry real cost. None ask for textbook definitions. Each demands a candidate demonstrate how they reason under constraints — measurement before action, trade-offs made explicit, individual ownership separated from team credit.
Take the React performance question. The expected answer starts with the profiler or Core Web Vitals, not a guess. It names specific failure modes — unnecessary re-renders, large bundles, blocking requests, unoptimized images, and explains how to confirm each. It finishes with a regression guard: a performance budget in CI. That signal selects for engineers who instrument first, optimize second, and automate the guardrail so the problem stays solved. The API spike question follows the same pattern: clarify read/write mix and consistency needs before designing, then layer caching, rate limiting, horizontal scaling, queueing, and database indexes. The candidate must say what they monitor and what degrades gracefully when limits hit. Systems thinking, not just coding, is the filter.
The end-to-end feature question forces clarity on individual contribution. "Be clear about what you did versus the team" appears in the rubric. Protex AI wants to hear the problem, the design choices, the trade-offs, the testing strategy, and a measurable result. Vague "we shipped" narratives fail. The model-shipping question separates offline metrics from business impact — baseline, held-out set matching production, error analysis on important segments, then an A/B test or shadow launch before full rollout. That is the voice of a company that has learned the hard way that a model that looks good on a test set can hurt customers in the field. The zero-downtime rollout question expects feature flags, canary or blue-green deploys, health checks, and automatic rollback triggers tied to specific metrics. Deployment safety is not optional; it is a competence the bar assumes.
These questions map directly to the skills the job board surfaces most often: AWS, AWS CloudWatch, Git, JavaScript, and CRM familiarity. Roles posted on Protex AI over the last 90 days — Senior DevOps Engineer, Engineering Manager, Infrastructure, Technical Solutions Engineer, and the go-to-market roles, all sit in a hybrid model the company describes as 100% hybrid. Seniority splits 50/50 between mid-level and entry-level, which means the bar must work for both a new grad who can demonstrate rigorous process on a capstone project and a senior hire who has run canary deployments at scale.
The careers page states the company hires "diverse, ambitious, and smart people, allowing them to do their best work every day." The interview rubric operationalizes that. Ambition shows up in the scope of the problems candidates are asked to design for — 10× traffic spikes, model rollouts that affect worker safety, performance budgets that survive CI. Smart shows up in the requirement to separate measurement from guesswork, offline metrics from business impact, individual contribution from team output. Diverse shows up in the range of roles open across engineering, DevOps, solutions, and enterprise sales, all remote-accessible from Dublin, Budapest, Boston, or Limerick.
What the bar does not select for is process compliance for its own sake. No evidence exists of algorithmic puzzles, whiteboard trivia, or multi-panel culture-fit screens. The signal: can you ship safely, measure honestly, and own the outcome? Candidates who need a spec handed to them, who optimize before they measure, who blur their role in a team win, or who treat model deployment as a one-time handoff will not pass. The company is small, roughly 60 people as of late 2026, and the $36 million Series B closed in 2024 (following seed in 2021 and Series A in 2022) buys runway, not bureaucracy. The hiring bar is the primary mechanism that keeps the pace honest.
Employee Feedback: Signal and Noise
Public employee feedback for Protex AI splits across two distinct identities, and the distinction matters for anyone evaluating the AI safety company. Glassdoor lists 12 reviews for Protex AI specifically, though the research digest captures no verbatim content from them. Meanwhile, Indeed hosts 16 reviews under "Protex the PT Xperts" (overall rating 3.3/5 as of July 2026), with detailed accounts spanning 2015–2026 that describe fleet management, field technician, and lab technician roles. Those functions do not appear on Protex AI's current job board, which lists engineering, sales, and customer-success openings in Dublin, Budapest, Boston, and remote US. The ThinkBusiness profile confirms Protex AI employs roughly 60 people and closed a $36 million Series B in 2024; the Indeed reviews describe a company with "technicians," "dispatch," and "hazardous trade" work. They are a separate entity sharing a name. Applicants should verify which organization a given review describes before weighing it.
The Indeed reviews attributable to "Protex the PT Xperts" show a polarized picture. Positive accounts emphasize a "family" atmosphere, supportive onboarding, and schedule flexibility. A 2022 reviewer wrote they felt "incredibly welcomed and encouraged by my Protex family" and noted the company was "open to teaching their incoming employees who may not have specific experience." Another from 2021 called it "family-oriented and fun" with "lots of work to be done but is fairly easy." A 2018 review praised certification opportunities and "work/home life balance." Negative reviews, concentrated in 2019–2021, cite low pay, broken promises, favoritism, and safety concerns. A 2019 reviewer called it "one of the worst companies in the trade" with "unlivable wage for the hazardous trade." A 2021 review alleged "favorites were played by supervisors" and "no 15 minute breaks were given just 30 minute lunch." A 2024 review described HR as "a joke they only come out when they need to get in trouble." A 2025 review claimed "Constantly under pressure to perform… even if you put in two weeks you don't get accrued PTO paid back." A 2026 review from a self-identified Fleet Manager said the role "was significantly different from what was presented during the hiring process" and expanded into facilities, safety, and construction labor.
Indeed's category scores (as of July 2026): Work-life balance 3.4, Culture 3.3, Management 3.0, Pay and benefits 2.9, Job security and advancement 2.9. The platform's summary tags list "Supportive environment," "Inclusive work environment," and "Ability to learn new things" under what people like; "Fair pay for job," "Sense of belonging," and "General feeling of work happiness" under areas for improvement.
None of these reviews map to Protex AI's current AI engineering, product, or go-to-market teams. The company's own careers page (protex.ai) and Zero G Talent's board show roles such as Senior DevOps Engineer (Budapest / Remote US), Engineering Manager, Infrastructure (Dublin / Remote US), and Enterprise Account Executive (Dublin / Remote US), none resemble the technician or fleet-manager positions in the Indeed corpus. Glassdoor's 12 Protex AI reviews remain the only direct signal for the AI safety business, but their content is not publicly excerpted in the research. Candidates should treat the Indeed archive as background noise and weigh Glassdoor's aggregate sentiment (unavailable in detail) against the documented hiring needs and funding trajectory of the AI company.
Who Thrives, Who Burns Out
The candidate who thrives at Protex AI treats ambiguity as a design constraint rather than a blocker. Open roles, Senior DevOps Engineer in Budapest, Engineering Manager, Infrastructure in Dublin, Technical Solutions Engineer across Dublin, Limerick, and US remote, signal a team building safety-critical computer vision systems that must deploy reliably across construction sites, warehouses, and manufacturing floors. These are not greenfield research posts; they demand production-grade judgment on model serving, infrastructure scaling, and customer-facing integration. Someone who has shipped ML systems to environments where downtime means halted operations or missed safety incidents will recognize the stakes. Enterprise Account Executive and Enterprise Account Executive roles in Dublin and Boston confirm the go-to-market motion targets large, regulated buyers, sales cycles are long, technical validation is rigorous, and the product must earn trust before the contract signs.
Founders Dan Hobbs and Ciarán O'Mara built a remote-first organization spanning Dublin, Budapest, Boston, and US remote hubs. That geography reflects a customer base across Europe and North America and a hiring strategy that pulls talent from both. The candidate who succeeds here has operated across time zones without a shared office to fall back on. They write decisions down, default to async communication, and treat documentation as a product artifact, not a chore. Global Demand Gen & ABM Lead and Enterprise Client Strategist roles suggest marketing and post-sales functions are still lean; the person who steps into those seats will define the playbook, not execute someone else's.
Burnout risk concentrates where autonomy meets accountability without guardrails. Glassdoor's 12 anonymous reviews, a small sample, but enough to surface a pattern at early-stage companies, show the same freedom that lets a high-agency operator move fast becomes a vacuum for someone who needs structure to prioritize. The Customer Success Manager role in Boston lists an hourly range of USD 8–12, which reads like a contractor or junior band; if that reflects the broader compensation philosophy for non-senior roles, candidates should calibrate expectations. The careers page promises "diverse, ambitious, and smart people" doing "their best work every day", language that, in practice, often translates to high output demands with limited process support.
The hiring bar selects for ownership signals: shipped products, owned incidents, led migrations, closed enterprise deals. The interview process tests for technical rigor and decision quality under incomplete information. A candidate who has only ever executed well-scoped tickets in a mature platform team will likely struggle. So will someone who equates "remote-first" with "flexible hours" without internalizing the discipline that makes async work, proactive communication, explicit handoffs, self-directed learning. The safety-critical domain adds a non-negotiable layer: model failures are not abstract metrics; they correlate to workplace injuries. That weight changes how you test, how you deploy, and how you respond to alerts.
Protex AI is not a place to learn how to be a senior engineer or a strategic account executive. It is a place to apply those capabilities at the edge of what computer vision can reliably do in industrial settings. The right candidate has already made the mistakes that teach judgment, elsewhere, and wants the latitude to apply it without asking permission. The wrong candidate mistakes latitude for lack of standards. The difference shows up in the shadow-mode logs, in the false-positive pipeline that the same engineers who ship the code own end to end.
Working in frontier tech? Zero G Talent tracks the openings: see every open Protex AI role, browse frontier tech jobs, the companies hiring, and the people building the field.