Inside the Fleet
Cobalt Robotics ships autonomous security robots to more than a hundred sites across five continents while running a 24/7 remote operations center — a combination that leaves no margin for choosing between speed and reliability. The fleet must patrol corporate offices for eight hours on a charge, the software stack pushes over-the-air updates to hundreds of robots weekly, and the command centers in California and Utah never go dark. The company grew from idea to paid deployments in twelve months, a pace set by founders Erik Schluntz, a former SpaceX engineer, and Travis Deyle, a roboticist from Google X and Georgia Tech's Healthcare Robotics Lab. That founding DNA — SpaceX's hardware urgency, Google X's moonshot tolerance — still shapes how decisions get made.
As of 2023, the team numbered about 200 people, with a core in Palo Alto and Orange County plus reserve engineers at client sites from Seattle to Manila. The reserve engineer role, listed on Zero G Talent's board, reveals the operational shape: engineers travel to customer sites, install and maintain robots, troubleshoot hardware in the field, and feed data back to the core team. It blends field service, devops, and customer engineering — there is no pure "lab" track. Schluntz stepped down as CTO in early 2024 after eight years, writing that the company had "succeeded in bringing robot security to over 100 sites on 5 continents and grown to a team of >150 people." Deyle remains CEO. The RaaS model means every robot is a recurring revenue line that demands uptime; a failed patch or a dead battery on a Friday night hits revenue directly.
The command center — staffed by operators hired largely from customer-service backgrounds, speaking two dozen languages — is the nervous system. When a robot flags an anomaly (an open door, a thermal spike, a badge mismatch), a specialist takes control via two-way video and audio, assesses, and escalates. That loop runs 24/7/365. Engineers there watch the same telemetry, push fixes, and coordinate with field engineers. The software stack — ROS, Python, HTTP APIs, NVIDIA 1080-class GPUs on board — is built for remote observability; over-the-air updates run on a weekly cadence.
The fleet sets the pace: hundreds of robots, eight-hour patrols, two-hour charges. That rhythm — hardware cadence, software velocity, operations urgency — defines a culture built on high ownership and fast iteration, the same culture that employee reviews now let candidates weigh against the mission pull.
The Stated Code
Cobalt publishes a short philosophy statement on its website: "We believe in empowering individuals to pursue their passions and curiosities in an environment built on mutual respect and maximum helpfulness." Mike LeBlanc, who co-founded the company and served as President and COO, distills his personal operating code into three lines: "Master Yourself; Master Your Craft; Master The World." Both come from the company and LeBlanc's own page; they frame the aspiration. Daily reality shows where the frame holds and where it bends.
The origin story doubles as an operating principle. In a 2019 talk, co-founder Erik Schluntz said he and Travis Deyle started Cobalt with no product idea. They interviewed people across industries hunting a problem, not knighting an idea. A physical-security professional described the gap between static cameras and human guards, which are worlds apart in cost and value. That conversation shaped the founding insight: a robot that patrols and video-calls a human specialist when something looks wrong. The principle of starting from the customer's problem, not your tech, drives the service model. Cobalt does not sell robots. It charges a weekly fee for a managed security service, like a guard company. "We are absolutely a security company where we provide security to our customers and we just happen to do that through a robot," Schluntz said in that same talk.
Human-robot collaboration drives the design. "Really this company is all about bringing people and robots to work together; these robots are really good at some things and people are really good at the others, and we strongly believe that by using both together we can create a solution that works best for everyone," Schluntz said. The robot patrols, detects anomalies, and initiates a video call. A remote specialist (Cobalt employs its own) handles the interaction. The robot never confronts. Schluntz emphasized that 99% of security guards are unarmed and trained to observe and report, not intervene. The robot follows the same procedure: deter, document, escalate.
That restraint drives the physical design. The team chose fabric covering because "it made the robot feel warm and comfortable and actually a little soft and squishy," Schluntz said. The goal: "we needed to make the robot feel comfortable and friendly to be around because if the robot scared people even if it was also scaring away bad guys it wasn't doing its job." Security means making employees feel safe, not chasing intruders. A robot spotted two women carrying a metal box out of a server-floor office at night; the specialist video-chatted them, learned they were removing burnt toast, and logged a harmless resolution. Another detected a slip-and-fall, alerted the specialist, who confirmed the employee was okay and offered to call medical help.
Speed and ownership drove early execution. The team went from first idea to first deployed robot in just over a year with only three engineers, "working very long hours working nights and weekends," Schluntz said. That pace set the culture: high ownership, fast iteration, hardware and software tightly coupled. The third surprise: "the cultural options and the human robot interaction it turned out to be very hard," a line that captures how much product work lives in behavioral nuance.
Reviewers praise flexible arrangements and camaraderie. They flag unclear direction, high turnover, inconsistent leadership, poor communication. Reviews name no executives beyond the leadership team. The pattern suggests the founding principles haven't consistently reached management practice. Candidates should weigh the gap between stated philosophy and reported experience.
Who Gets Hired
Cobalt's hiring signal has stayed consistent since the founding team of three engineers shipped a deployed robot in just over a year. The founder who described that sprint, working nights and weekends to bridge hardware, software, and the "unexpectedly hard" human-robot interaction layer, explicitly tied the pace to the caliber of people they recruited: "our focus on getting really top tier engineers that have worked on very interesting technology machine learning in robotics in human robot interaction before I really helps us do that." That statement, captured in a 2019 talk, still matches the roles on the job board today.
The interview process is thin by design; this section does not reconstruct stages or timelines. What surfaces from public accounts is a pattern: candidates describe a clear, organized process with relevant questions focused on skills and a welcoming atmosphere emphasizing effective communication. Glassdoor shows 16 interview reviews and 19 posted questions; Indeed aggregates 18 Q&A entries. The consistency of those descriptors (skills-focused, communicative, organized) suggests the filter is deliberate, not improvised. The founders "went out and interviewed lots of different people looking for some sort of opportunity some sort of problem that we could solve using the technology that we knew," Schluntz said. They built the product around a discovered problem, not a preconceived idea, and the hiring bar reflects that same empirical bias; they select for engineers who have already navigated messy, cross-disciplinary robotics challenges.
The "third part" the founder flagged, cultural options and human-robot interaction, is a hiring signal in itself. Cobalt partnered with Fuseproject to design a robot wrapped in stretched fabric rather than plastic or metal, explicitly to make it feel "warm and comfortable" in office lobbies. That choice means every engineer touches the product's social interface, not just its navigation stack. The Reserve Engineer role, distributed across five countries, runs the fleet that customers now demand faster than production can deliver. The hiring bar therefore weighs deployment reliability and customer-facing judgment alongside core robotics chops. Candidates who have shipped hardware that lives in occupied spaces, such as hospitals, offices, and data centers, carry an edge. So do those who have worked in service-model robotics where uptime and remote support are product features, not afterthoughts.
The research shows no published competency framework or list of target schools. It shows a company that started with three engineers, scaled to five continents, and still describes its interview process in terms candidates recognize as skill-focused and communicative. The traits that predict an offer include prior work at the ML/robotics/HRI intersection, comfort with hardware-software integration in the field, and a track record of shipping into environments where non-technical people interact with the machine daily. The board's current openings, the founder's own hiring rationale, and the public interview reviews point to that profile.
Reading the Reviews
About 75 anonymous employee reviews for Cobalt Robotics sit on Glassdoor, with a comparable pool on Indeed; this data set is large enough to spot patterns but small enough that any single review carries outsized weight. The research available to this article lacks the full text of those reviews, their star ratings, or their dates. That absence matters: the best way to assess culture from the outside is to read the reviews directly rather than rely on a summary.
The review count tells you that enough current and former employees have written publicly to surface recurring themes. In robotics companies of this size, roughly 150–200 people across multiple sites, a review base in the high double digits usually means a workforce that feels strongly enough to post, whether positively or negatively. Treat the Glassdoor and Indeed pages as primary sources and look for three things: consistency of complaint or praise across multiple reviewers, recency (reviews from the last 12 months weigh more than older ones), and whether the reviewer's role matches the one you're targeting. A software engineer's experience of sprint cadence and code-review culture differs sharply from a field operations technician's experience of travel schedules and on-site escalations.
| Role | Compensation | Source |
|---|---|---|
| Reserve Engineer (field) | $20–$50/hr | Zero G Talent board |
| Salaried roles (9 titles) | $42k–$104k/yr (median six figures) | Job board |
The board shows Cobalt hiring for Reserve Engineer roles in Seattle, San Francisco, Memphis, London, and Melbourne, plus a Solutions Engineer role in Manila at 550 to 760 Philippine pesos an hour. Those postings confirm the geographic spread reviewers are likely describing. If multiple reviews from different locations cite the same friction, say communication latency between headquarters and field teams, that pattern carries more weight than a single-location grievance. Conversely, if praise for autonomy or ownership appears across engineering, operations, and support roles, it signals the founder-driven culture described elsewhere in this article reaches the organization rather than staying siloed in one team.
Cross-reference review themes with the company's public operating principles. When reviewers use language that echoes stated values, such as mutual respect, maximum helpfulness, and mastering your craft, it shows the principles live, not laminate. When the language clashes, with reviewers describing lack of clear direction, high turnover, or inconsistent leadership, the gap between aspiration and reality becomes an interview question: ask for a recent example of how the team resolved a priority conflict between shipping a feature and fixing a field issue.
The board's salary band across nine salaried roles anchors compensation for evaluating reviews that mention pay. If multiple reviewers call compensation "below market" while the median sits at six figures, the disconnect may stem from role-specific bands, equity expectations, or cost-of-living pressure in the Bay Area and London. Weigh those reviews against the specific role and location.
Bottom line: the review volume is there. The content sits on Glassdoor and Indeed. Sort by role and date, and test the culture claims against the patterns you find in the interview loop.
Who Stays, Who Leaves
The profile that succeeds at Cobalt reads like the inverse of the friction points that show up in public reviews. Reviewers consistently cite flexible arrangements and team camaraderie as positives, while flagging those same friction points and poor communication as negatives. That pattern selects for engineers and operators who treat ambiguity as a design constraint rather than a blocker. People who have shipped hardware in early-stage robotics companies, or who have run field deployments where the spec changes the day before install, tend to map to the "maximum helpfulness" value the founders codified on the careers page. They show up, find the gap, and close it without waiting for a ticket.
The Reserve Engineer roles posted across five countries signal the operational shape of the team: distributed, field-facing, expected to operate with minimal oversight. A candidate who has managed a robot fleet across time zones, debugged a perception stack on a customer site, or written the runbook the next hire will follow will recognize the pace. The Solutions Engineer posting at the Manila rate alongside U.S. Reserve roles at the U.S. rate also reveals a cost structure that relies on high-leverage individual contributors rather than layers of management. Someone who needs a defined career ladder, regular 1:1s with a dedicated manager, or a predictable sprint cadence will likely churn inside six months, exactly the turnover the reviews describe.
Burnout clusters around three mismatches. First, the communication gap: reviewers describe leadership as inconsistent and direction as unclear. A person who relies on top-down prioritization, detailed requirements docs, or a stable product roadmap will spend more energy chasing clarity than building. Second, the ownership ceiling: "maximum helpfulness" works when everyone pulls sideways; it fractures when someone waits for permission. Third, the mission-intensity fit. Cobalt sells autonomous security robots to enterprises that expect high uptime. Field incidents are not hypothetical. Engineers who joined for the robotics problem but did not sign up for the customer-facing reliability burden, such as site escalations and the physical wear of travel, tend to exit. The median board salary at that level reflects a market rate for that blend of depth and grit, not a premium for comfort.
Self-assessment is straightforward. If your best work happened when the spec was vague and you built the test rig, the CI pipeline, and the handoff doc yourself, you will likely thrive. If the opposite describes your experience, the culture will feel like a vacuum. The reviews are not subtle about this. Read the last twenty Glassdoor entries; the people who stay describe autonomy, impact, and team. The people who leave describe chaos, silos, and burnout. Both groups are describing the same environment: a fleet that never sleeps, wrapped in fabric, waiting for the next anomaly to appear on a screen in California or Utah, where a specialist answers the call and the robot rolls on.
Working in robotics? Zero G Talent tracks the openings: see every open Cobalt Robotics role, browse robotics jobs, the companies hiring, and the people building the field.