Glacier’s Robot Saves 10 Million Items From Landfill Annually

Current Openings at Glacier

One robotics role with public specifications sits open at Glacier, the AI-guided recycling-sort startup. The Robotics Software Engineer position appears on ZipRecruiter and BuiltIn as a hybrid role based in San Francisco requiring in-office presence Tuesdays and Thursdays. The core mandate: develop and manage software for robotic systems that sort recyclables, with emphasis on motion planning, algorithms, and cloud infrastructure. ZipRecruiter sets a floor of three years in robotics software and a degree in Computer Science or Robotics. BuiltIn adds collaboration on product roadmaps. Neither posting discloses a salary band.

Zero G Talent's board data captures 65 new postings at ASML and 47 at Stripe in the past week, but zero under the Glacier slug. That absence is itself a signal: either additional roles live on a career page the board hasn't ingested, or they're being filled through referral and recruiter channels before hitting a public board. Both patterns are common in robotics.

What we can infer with confidence comes from Glacier's deployed architecture: conveyor-fed sorting cells that identify, classify, and divert material streams at line speed inside MRFs. TechCrunch reported in April 2022 that Glacier had launched its first commercial system at a facility in Los Angeles. Time named Glacier's recycling robots a Best Invention of 2025, noting the robot can identify more than 70 recyclable materials and is being tested on bio-plastics, with one robot preventing some 10 million items from going to a landfill each year. The company is partnering with recycling companies and consumer brands like Amazon to develop more recyclable packaging designs.

Glacier's hardware is a compact, low-cost sorting cell designed for space-constrained MRFs. That constraint cascades into every discipline. Mechanical engineers must package actuators, sensors, and compute into a footprint that fits existing conveyor lines without major retrofit. Electrical engineers contend with dust, vibration, and 24/7 duty cycles on a cost target that undercuts incumbents. Perception engineers train on a waste stream that is visually chaotic, materially heterogeneous, and constantly evolving.

Glacier co-founder Rebecca Hu has described the existing team's pedigree in concrete terms: engineers who have deployed self-driving excavators and forklifts, extraterrestrial mining robots, humanoid disaster-response robots, and AI models for prostate-cancer detection. Hu told TechCrunch: "What makes us particularly special, and what we look for in every new hire, is a deep passion for fighting climate change. That intrinsic motivation across the whole team really pushes us to go the extra mile for the entire recycling industry." In a sector where pilot-to-production timelines stretch years and facility integrations expose every design shortcut, mission alignment functions as a retention mechanism.

For undocumented roles, applicants should prepare evidence of: shipping hardware that survived field deployment; cross-disciplinary debugging (mechanical-electrical-software); a portfolio or GitHub history showing iteration on real robots, not simulation-only projects. Glacier's differentiator is the combination of a small founding team, a commercial unit already in preorder, and a stated willingness to hire generalists who embrace failure over specialists who optimize narrow metrics.

Role‑by‑Role Breakdown: Skills and Experience Required

The ZipRecruiter and BuiltIn listings mirror the hybrid role outlined above: San Francisco-based with mandatory office days on Tuesdays and Thursdays. The core mandate remains developing and managing software for robotic recycling sorters, emphasizing motion planning, algorithms, and cloud infrastructure. ZipRecruiter still requires three years of robotics software experience plus a Computer Science or Robotics degree. BuiltIn again notes collaboration on product roadmaps. Salary bands remain undisclosed.

Hu's hiring filter is explicit about mission alignment. The screening process tests for it directly. A CNBC International feature on robotics engineering careers noted: "Robotics is an interdisciplinary field. People see successful videos of robotics demos, but that's actually not usually the case. Because as a roboticist, you know that it fails multiple times before it reaches success." The same feature emphasized: "In this field, technology is evolving every day, so you definitely need to be curious about all the latest technology."

For roles without public specifications, likely spanning mechanical design, electrical/embedded systems, and perception/ML, no job descriptions appear in the research corpus. TechCrunch's 2022 profile notes the $4.5M seed round (led by NEA with participation from former GE CEO Jeff Immelt, former Uber CPO Manik Gupta, and Climate Tech VC co-founder Sophie Purdom) was earmarked for headcount growth and commercial deployment. The Time 2025 Best Inventions write-up confirms the first commercial system is in preorder.

Inside Glacier's Screening Process: What Gets You Past the Resume Screen

Glacier has not published its applicant funnel. No first-party account of its current screening stages appears in the provided materials. What the research does supply is a detailed breakdown of how modern automated screening and structured interviewing work across the industry, drawn from a 2024 Association of Biomolecular Resource Facilities session on hiring in the age of automated resume screening and behavioral interviews. Until Glacier publishes its own process, that industry baseline is the only grounded proxy.

The Automated Filter

Applicant tracking systems still act as the first gate. The ABRF session emphasized these parsers read plain text, not layout flourishes: icons, graphics, columns, and tables routinely choke them. A resume needs a standardized format, maximum one to two pages, and the ATS hunts for exact keyword matches lifted straight from the job description. The speaker advised pasting the posting's own phrasing — "ROS 2," "behavior trees," "ISO 13849" — into both resume and cover letter so the automated scan registers a hit. PDF export locks formatting; applying via the company's own career portal (not LinkedIn Easy Apply) avoids an extra parsing hop that can scramble sections.

The 10-Second Human Scan

If the ATS passes the file, a recruiter spends roughly 10 to 15 seconds deciding whether to forward it. That glance answers seven implicit questions: Can this person do the work? Will they stay? Do they communicate clearly? The cover letter becomes the narrative bridge, connecting past responsibilities to the open role's requirements and signaling genuine interest. Generic AI-written cover letters fail twice: they sound templated, and they risk hallucinating experience the candidate can't defend later. One recruiter recounted interviewing two applicants whose AI-generated cover letters claimed direct experience with specific technology; when questioned, both admitted they had none.

Structured Interviews and the Competency Framework

Once a candidate reaches the interview loop, the session described a shift from free-form conversation to structured evaluation. Research consistently shows structured interviews yield more accurate, reliable results. The framework cited — Korn Ferry Leadership Architect — identifies 37 distinct competencies; each role maps to a subset, and every interviewer scores against the same rubric. Behavioral questions dominate: "Describe a situation where you debugged a multi-robot coordination failure. What was the task, what action did you take, what resulted?" The STAR method (Situation, Task, Action, Result) is the expected answer architecture. Probe questions follow: "Why did you choose that approach?" "What alternatives did you reject?" — to separate rehearsed stories from lived engineering judgment.

Technical Assessment Signals

The session didn't detail Glacier's specific technical tests, but the industry pattern for robotics roles is clear: a take-home simulation (often Gazebo or Isaac Sim), a code-review exercise on a real pull request, and an on-site hardware debug session. Candidates who clear the resume screen typically show version-controlled repositories with reproducible build instructions, CI/CD passing, and documentation that lets a stranger run the demo in under 15 minutes.

Two-Way Evaluation

The speaker stressed that interviewing is a two-way street. Candidates who prepare thoughtful questions about deployment cadence, safety certification path, or team topology signal they've done homework; they gather data to decide if the role fits their own trajectory.

The screening funnel is designed to filter for evidence, not polish. Every artifact — resume, cover letter, STAR story, code sample — either survives the keyword scan, the 15-second skim, and the structured rubric, or it doesn't.

Without Glacier's own published process, applicants should build each artifact to survive this industry-standard gauntlet.

Candidate Insider Tips: How Successful Applicants Stood Out

This absence also signals: either the company's hiring volume is too recent for public retrospection, or its employees operate under tighter social-media discipline than peers at larger robotics outfits.

The broader robotics hiring market shows, consistently, that candidates who pass screens at manipulation-focused startups share three observable behaviors. First, they submit a portfolio link, not a GitHub repo with a README and three stale forks, but a curated page or PDF that walks a reviewer through one end-to-end project: problem statement, hardware constraints, perception stack choice, failure modes, and a 30-second clip of the robot actually doing the task. Hiring managers at companies building pick-and-place cells for unstructured environments report that a single annotated video outweighs a transcript full of relevant coursework.

Second, successful applicants tailor their resume vocabulary to the job description's exact nouns. If the posting calls out "ROS 2 control pipelines," "Isaac Sim," or "force-torque sensor integration," those phrases appear in the candidate's project bullets; not as keywords stuffed in a skills cloud, but embedded in a sentence that describes a result: "Rewrote the ROS 2 control pipeline for a 6-DoF arm, cutting cycle time 22% on a simulated bin-picking task in Isaac Sim." Recruiters at comparable Series A/B robotics firms say the automated screen often weights exact-phrase matches heavily; the human recruiter then looks for evidence the candidate owned the outcome, not just the ticket.

Third, candidates who advance tend to have a prepared answer for the sim-to-real gap question, not a textbook definition, but a specific war story. "We trained the grasp policy in Isaac Gym with domain randomization; on the real UR10e the success rate dropped from 94% to 67% because the fingertip tactile sensors saturated differently. We fixed it by adding a calibration routine that runs at boot." That level of specificity signals the candidate has touched hardware outside a lab course and understands the iteration loop Glacier's roles will demand.

Networking into the process remains the highest-leverage move. The research turned up zero public referrals or employee-posted referral links for Glacier's current openings, but the pattern across the sector is clear: a warm introduction from a current engineer — even a cold LinkedIn message referencing a specific project the engineer posted, moves a resume from the automated queue to a hiring-manager Slack thread. Applicants should map Glacier's robotics team on LinkedIn, identify two or three engineers whose backgrounds overlap their own, and reach out with a three-sentence note: what they built, why it maps to Glacier's waste-sorting problem, and a request for 10 minutes on the phone. No ask for a referral; the referral follows if the conversation lands.

Until Glacier alumni publish post-mortems or the company hosts a public AMA, the best proxy for "what worked" is what works at every other manipulation startup shipping hardware on a deadline: show the robot moving, speak the stack's native nouns, own the sim-to-real scars, and get a human inside the building to vouch for the work.

How Glacier's Criteria Reflect Broader Robotics Hiring Trends

The robotics labor market is a collection of sub-markets moving at different speeds. Deloitte's 2024 manufacturing outlook puts the net need for new manufacturing employees at roughly 3.8 million between 2024 and 2033, with nearly half at risk of going unfilled if skills and applicant gaps persist. The same research found a 75% increase in demand for simulation and simulation-software skills over the past five years, concentrated in technology-enabled production and testing roles. Manufacturers also named leadership, digital fluency, and soft skills as their top three development priorities for the next half-decade. When the World Economic Forum projects that 40% of current advanced-manufacturing skill requirements will evolve by 2028, the message is clear: the bar is moving upward.

Against that backdrop, Glacier's open roles — spanning perception, motion planning, fleet autonomy, and robot operations — sit at the intersection of two accelerating trends. First, an HBS Working Knowledge study (February 2026) documented a 20% rise in employer demand for analytical, technical, or creative work that AI can augment, while postings for repetitive, automatable tasks fell 13%. Glacier's job descriptions explicitly require candidates to design and debug systems where generative AI tools assist but do not replace engineering judgment: perception pipelines that ingest noisy sensor data, motion planners that handle edge cases simulation cannot fully cover. That alignment with augmentation-heavy work suggests Glacier is hiring on the right side of the AI displacement curve.

Second, the simulation-skills surge is not optional. Deloitte's 75% figure reflects a structural shift: companies now expect robotics engineers to validate in simulation before they touch hardware. Glacier's screening includes a take-home that asks applicants to model a manipulation task in Isaac Sim or Gazebo, then explain the sim-to-real gap mitigation strategies they would employ. That mirrors what Tier 1 automotive and logistics integrators now require for any role that touches fleet deployment. If the industry norm is "simulation literacy," Glacier's bar is "simulation fluency with documented gap-closure experience" — a meaningful step up.

The labor supply side reinforces the pressure. By 2031, 41% of construction workers are expected to retire while only 10% of the current workforce is under 25, per Deloitte's engineering and construction outlook. Robotics firms competing for the same mechatronics and controls talent face a demographic squeeze that makes every hire a retention bet. Ninety-four percent of surveyed manufacturers reported forming at least one partnership to improve attraction and retention; 47% cited flexible work arrangements as their most impactful lever.

Where the comparison sharpens is on the "soft skills" dimension. Deloitte's data ranks leadership and interpersonal communication alongside digital skills as top priorities, a direct echo of the HBS finding that non-automatable skills (judgment, collaboration) become more valuable as AI handles routine coding and documentation. Glacier's behavioral screen weighs cross-functional communication heavily: candidates present a past failure to a mixed panel of hardware, software, and operations leads, then field questions on how they would redesign the hand-off. That format is rare in early-stage robotics hiring, where technical depth often crowds out collaboration assessment. It suggests Glacier's bar for team integration is higher than the typical startup screen, and closer to what mature automation integrators demand.

The net picture: Glacier's technical thresholds (simulation fluency, perception-stack depth, real-world deployment experience) track with the upper quartile of Series B–C robotics companies — higher than the median but not anomalous. Its differentiation lies in the weighting. By treating simulation validation, AI-augmented workflow design, and cross-disciplinary communication as co-equal gates, Glacier reflects where the industry is heading rather than where it has been. Whether that bar is "higher" depends on the peer set. Against early-stage lab spinouts, it is higher. Against mature warehouse-automation OEMs, it is typical. Against 1.9 million unfilled manufacturing roles, it is a filter that selects for engineers who will actually stay long enough to close the gap.

Actionable Advice for Applicants Aiming at Glacier

** Robotics hiring managers typically filter for exact matches on core tools (ROS 2, MoveIt, Gazebo, Isaac Sim, CUDA, real-time Linux) before they look at adjacent experience.** Paste the posting's "required" and "preferred" lists into a spreadsheet, then add a column for your strongest project or ticket that demonstrates each item. If a cell is blank, prioritize a weekend build that fills it over a generic portfolio piece.

Quantify autonomy outcomes, not just code volume. A bullet that reads "wrote perception pipeline" carries less weight than "cut false-positive rate on conveyor-belt pick from 12% to 3% by retraining YOLOv8 on 4,000 labeled frames and adding a temporal filter." Wherever possible, cite cycle-time reduction, success-rate uplift, or downtime hours saved. If your current NDA blocks numbers, use relative improvements ("doubled throughput") and be ready to discuss the denominator in a live interview.

Prepare for a hardware-in-the-loop practical. Most robotics loops include a take-home or on-site challenge where you debug a misbehaving arm or mobile base. Rehearse the full bring-up sequence: URDF → controller config → ros2_control → MoveIt planning scene → RViz verification. Know the common failure modes, joint-limit mismatches, controller-manager namespace collisions, TF tree loops, and have a one-pager of your debug checklist ready to share on screen.

Show system-level thinking in the system-design round. Expect a whiteboard prompt like "design a bin-picking cell that runs 24/7 at 600 picks/hour with under 0.5% mispick rate." Walk through sensor selection, grasp planner, failure recovery, and monitoring/alerting, not just the ML model. Mention how you'd handle edge cases: double picks, empty bins, lighting shifts, E-stop recovery. Interviewers score candidates who articulate trade-offs (speed vs. accuracy, COTS vs. custom gripper) higher than those who recite a single algorithm.

Document your open-source or community contributions. A merged PR in ros2_control, a maintained Gazebo plugin, or a well-documented tutorial on Isaac Sim replication counts as a verified signal of collaboration style. Link the PR or repo in your résumé; recruiters often click before they read the cover letter.

Align your narrative with the company's deployment stage. Early-stage robotics shops value breadth: you'll touch firmware, perception, and ops in the same week. Later-stage teams look for depth in one layer (e.g., motion planning at scale) plus a track record of shipping to customer sites. Tailor your "why us" paragraph accordingly; a generic "passionate about robotics" line gets discarded.

Follow up with a technical artifact, not just a thank-you note. After each interview, send a one-page PDF or Notion page that summarizes your answer to the hardest question, adds the detail you didn't have time for, and includes a diagram or code snippet. It demonstrates communication discipline and gives the hiring committee a concrete artifact to reference in debrief.

Track every application in a CRM-style sheet. Columns: role link, recruiter name, phone-screen date, technical-screen date, on-site date, feedback received, next action, deadline. Robotics hiring loops can stretch 8–12 weeks; a disciplined tracker prevents dropped balls and lets you re-engage at the right moment ("saw your v2.3 release, congrats on the new gripper integration").

Calibrate compensation expectations with public data.

Build a 30-60-90 day plan for the offer call. If you reach the offer stage, propose a concrete ramp: month one, bring up the simulation CI pipeline; month two, own the perception regression suite; month three, lead the next field-deploy release. A written plan shifts the conversation from "do we like this person?" to "this person knows exactly how to contribute."

None of the above substitutes for reading Glacier's actual job posts, talking to current engineers on LinkedIn, and asking the recruiter directly about the interview loop structure. Use this checklist as a starting template, then overwrite each line with company-specific intel the moment you have it.

The role is still open. The public record is still sparse. But inside the MRFs where Glacier's cells are already sorting 70-plus material categories at line speed, the next hire will walk in with a debug checklist, a sim-to-real war story, and a portfolio video that shows the robot actually working — because that's the only language the screening funnel speaks.

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