Cartage Closes $200K ARR in Just Three Hours

The Five Open Roles at Cartage

Cartage, a Summer 2024 Y Combinator company, closed a $3.3 million seed round in October and runs a 12-person team after acquiring Westcore Logistics, Canada's fastest-growing logistics company in 2023. The acquisition brought COO Harman Sahota, who scaled Westcore from zero to $50 million in revenue in four years. Since pivoting from selling software to freight brokers to replacing them entirely, Cartage has signed eight shippers — from snack startups to the second-largest U.S. furniture manufacturer, and hit an inflection point the founders describe bluntly: $200,000 in ARR closed in three hours. The company is on pace to coordinate more than $2 million in freight over the next six months.

That velocity drives five open roles listed on Y Combinator: GTM Engineer, Autonomy Manager, Autonomy Operations Architect, Full Stack Engineer, and Sales Development Representative. Together they form a coherent picture: Cartage bets that technical leverage, not headcount, will scale its top of funnel and its product simultaneously.

The GTM Engineer role is the most unusual. The posting frames the problem directly: "We're great at closing deals but our top of funnel relies on our small sales team doing outreach. We want to scale our top of funnel using technical leverage, not mass hiring." The hire must "help Wilson sell himself," Wilson being Cartage's AI agent that already sends emails, texts, and calls to coordinate freight. The mandate: build automated outreach from the ground up, turning the same agentic capabilities that move freight into a prospecting engine. This is not a growth marketing role. It is a product engineering role disguised as go-to-market, signaling that Cartage treats distribution as a software problem.

The SDR role exists in parallel: "proactively identify and create new sales opportunities through high-volume cold calling" with a quota of 10-plus qualified meetings a month. The coexistence of an SDR and a GTM Engineer is telling. The SDR handles human edge cases and relationship nuance the agent can't yet manage; the GTM Engineer builds the system that eventually absorbs the repeatable portions of that work. Cartage runs both playbooks at once, using human reps to validate the scripts the agent will later execute at scale.

The Full Stack Engineer role targets the product core: "Build autonomous agents that can reliably automate complex, data-intensive workflows, handling everything from quoting to tracking." The posting emphasizes two infrastructure priorities rare in early-stage specs: automated AI evaluation loops ("monitor, evaluate, and retrain AI models in production, ensuring performance improves automatically as new edge cases arise") and deep integrations across enterprise systems (SAP, Oracle, ERPs, TMSs) plus messaging platforms (Slack, Teams) and browser automation. This is not feature work. It is reliability work, the plumbing that determines whether an AI agent can be trusted with a $2 million freight book without human supervision.

The Autonomy Manager and Autonomy Operations Architect roles round out the team, focusing on the operational side of Cartage's autonomous freight coordination.

Together, the five roles map to a single thesis: Cartage is building a service-as-software system where the marginal cost of coordinating an additional shipment approaches zero. The GTM Engineer lowers the marginal cost of acquiring a customer. The Full Stack Engineer lowers the marginal cost of serving one. The SDR bridges the gap until the first two succeed. The autonomy roles manage the operational layer.

Cartage's founders — CEO Abdul Basharat (formerly product lead for network and PLG at Rose Rocket, YC S16) and CTO Josh Lampen (founding engineer on Rose Rocket's Platform team), know the freight ERP landscape from the inside. They chose to build Wilson rather than sell into it. The hiring plan confirms they're still in the build phase, but the revenue inflection means they've entered the prove-it phase. The five roles are the team they need to prove Wilson can coordinate freight cheaper, faster, and more transparently than the broker ecosystem they're targeting.

Inside the Application Screen

Cartage's career page is hosted on Y Combinator's workatastartup.com platform — standard for a YC S24 company, but the volume hitting that pipeline is anything but standard. Scoutify, which monitors Cartage's career page directly, detects new postings within minutes and funnels applications through its own auto-apply layer across 50,000-plus companies. That infrastructure means every role Cartage posts receives a surge of machine-submitted resumes before a human ever logs in.

Scoutify's data shows three roles posted on May 13, 2026: GTM Engineer, Sales Development Representative, and Full Stack Engineer.

The posted base range acts as a coarse filter, senior engineers expecting typical mature enterprise SaaS rates self-select out or get flagged as compensation mismatches. Cartage lists San Francisco and Vancouver as its two hubs, with three roles in San Francisco and two in Vancouver, so candidates without work authorization in either country drop out automatically.

Beyond the ATS, Cartage's own product — Wilson, an AI agent that manages freight end-to-end from [email protected], signals how the company thinks about automation. A startup building an AI to replace repetitive coordination work applies that same lens to its hiring funnel. The most-hired roles over the last 90 days mirror the current openings: one Full Stack Engineer, one SDR, one Autonomy Operations Architect. That symmetry confirms the screening criteria have stabilized around a repeatable profile.

Scoutify's candidate-facing tooling reveals the next layer. The platform collects "real questions reported by candidates plus likely questions inferred from Cartage's open roles," a dataset built from actual interview loops. Those inferred questions become the de facto study guide for the automated screen. The GTM Engineer posting emphasizes "production systems at scale," a phrase Cartage uses in its own marketing.

Cartage has no public H-1B petition history, and its visa policy is unstated. International candidates in the Vancouver pipeline face Canadian work-permit requirements; the San Francisco pipeline follows standard U.S. work-authorization logic.

The result is a funnel that rewards specificity. The screen does not read for potential; it reads for signal match. Candidates who treat the job description as a specification document — mapping each requirement to a concrete artifact, clear the bot. The rest join the majority that never reach a human.

Passing the Bot: Keywords, Portfolios, and Referrals

Cartage's application volume has turned its automated screen into the primary gatekeeper. Industry data shows 52% of large logistics companies already leverage AI tools in recruitment, a figure expected to rise 30% in the next few years. At Cartage, that layer is not optional. If your résumé does not clear the filter, your application dies silently with no rejection email and no explanation.

The scoring engine reads for keyword matches, formatting consistency, relevant experience, and red flags like unexplained gaps. Parsers weight earlier content more heavily, so the highest-scoring material must sit on page one. A clean, single-column layout in PDF or .docx using Arial, Calibri, Georgia, or Times New Roman is the baseline. Headers and footers are ignored entirely, contact information belongs in the document body. Images, logos, and icons are invisible to text parsers and waste scoreable space. Consistent date formatting matters: mixing "Jan 2023," "January 2023," and "01/2023" in the same document confuses timeline parsing.

Keywords must appear in context, not as stuffed lists. The job description's exact phrasing should mirror the language in your experience bullets. Quantified achievements outperform vague claims: "reduced dwell time 18% across 12 facilities" beats "improved operational efficiency." Standard section headings — Experience, Skills, Education, help the parser map content to the rubric. One to two pages is the sweet spot; longer résumés are not penalized but the first page carries disproportionate weight.

For technical roles, a GitHub portfolio or project appendix functions as a secondary proof layer. The screen looks for concrete, scoreable credentials. Prior experience with AI-related tasks, even informal, registers as signal. "Completed 200+ RLHF evaluation tasks on Mercor" is a credential generic logistics résumés lack. Domain expertise leads: if you hold a CDL, hazmat endorsement, or TSA security clearance, that belongs in the first bullet, not buried at the bottom.

Referrals bypass the queue. Candidates with non-traditional paths — freelance project work, career pivots, fragmented employment, should prioritize referrals because AI screening is rigid and struggles with candidates who do not fit a standard mold. Opting out of automated screening is not available on most platforms; for AI training platforms like Mercor, DataAnnotation, and Outlier AI, the process is fully automated with no opt-out option. Your only path through is a CV the system can read and score correctly.

From Screen to Interview: The Human Layer

The automated screen is a gate, not a verdict. Candidates who clear the initial filter enter a human evaluation process that, at early-stage autonomous logistics companies, follows patterns established by companies moving physical assets at scale. Research on Cartage's specific interview loop is thin; Indeed's candidate portal lists generic preparation advice but no detailed breakdown. What exists is a consistent template across the frontier tech cohort: a funnel that narrows fast, weights demonstrated ownership over credential signaling, and tests whether an engineer can operate inside ambiguity.

The first human touchpoint is typically a recruiter screen, 30 minutes, behavioral, verifying the narrative the resume and AI scan already surfaced. Uber's engineering blog described this step as listening for specificity: "I really like to see words and phrases that demonstrated how much candidates owned in their prior roles." Vague answers get flagged.

Next comes a technical conversation with a hiring manager or senior engineer. At Carta, this was a single engineering chat after the recruiter screen. At Uber, a dedicated technical screen before the onsite. The signal is consistent: practical problem-solving over algorithmic recall. A 2021 Carta interviewee noted "the questions being asked were not esoteric... practical problems that required you to work together towards solutions. The focus was more on turning ambiguity into specific requirements, iterating and making sure you covered all reqs, not testing if you knew specific algorithms." That mirrors what Uber's recruiters described: engineers who "can break down what they've done and explain it, and also see how their work fits into the bigger picture."

System design follows. The 2026 interview-cycle data shows three recurring prompts at mid-level: design a banking system, design a cloud storage system, design a data pipeline. The evaluation rubric rewards decomposition: identifying consistency requirements, partitioning strategy, failure domains, observability.

Object-oriented design appears more often than LeetCode hard problems. The same 2026 cycle surfaced elevator systems, Minesweeper, banking applications, exercises that expose class structure, SOLID principles, and day-to-day modeling judgment. "They were focusing a little bit more on your understanding of class structure... versus like the brain-teaser LeetCode problem style," a candidate reported.

Behavioral rounds are competency-based and weighted toward negative-experience narratives. The 2026 trend: "tell me a time you received difficult feedback, tell me a time you disagreed with someone, tell me a time that you balanced competing priorities." Follow-up pressure is standard: "Okay, that's great. Tell me another time this happened." Arup's assessment centre model maps directly: values questions framed as hypotheticals, leadership and client-relationship probes, group exercises scored for inclusive collaboration. Uber's recruiters put it bluntly: "Are they a collaborator?... are they actually committed to their work, or are they clocking in and clocking out?"

The onsite — virtual or physical, compresses these into a single day. Uber's 2015 model: five hours, two behavioral, two technical (coding + system design), a lunch interview, a hiring-committee manager interview. Carta's was lighter: technical pair programming, behavioral culture-fit. The constant is cross-functional exposure. Candidates meet the people they'd ship with. At Arup, existing employees chat with candidates and "they'll probably be asked to give feedback on what they thought of you."

Final decisions tilt toward the last conversation. Arup's directors "put more weight on the final interview." Uber's hiring committee reviews the full packet but the manager interview often resolves borderline cases. For Cartage, with five roles open and a surge of post-screen applicants, the human layer is where the funnel tightens. The AI screen bought throughput. The interview loop buys signal — ownership, judgment, collaboration, that no classifier reliably extracts.

What This Signals for Frontier Tech Hiring

Cartage's five-role sprint and the AI screen sitting in front of it are the leading edge of how early-stage autonomous logistics companies now hire. The pattern repeats across the sector. Migrate Mate tracked 138 live open roles across seven logistics startups as of October 2, 2026, and the velocity at those companies mirrors what Microsoft's Work Trend Index calls "Frontier Firms": organizations where 95% of leaders say they are hiring for AI-specific roles, compared with 78% globally. On LinkedIn, the most prominent startups have grown headcount 20.6% year over year, nearly twice Big Tech's 10.6% pace.

The economics behind that growth explain the screening pressure. One-third of frontier-tech fundraising dollars went to hardware-focused VC funds in the most recent cycle, the highest share in a decade and up from 20% in 2021, per SVB. Capital intensity is rising — Stanford's 2025 AI Index shows training compute for notable models doubling every five months, and autonomous freight startups sit at the intersection of that compute demand and the physical-world deployment it requires.

The "fewer than five candidates per position" profile is exactly where HackerEarth's analysis says AI hiring tools underperform: public skill signals are sparse, and heavy automation measurably weakens employer brand with candidates who have options.

The bias profile is different from manual screening, not absent. The EEOC's disparate-impact framework and the EU AI Act's high-risk classification for employment AI both now require vendors to provide auditable explainability reports. Amazon's 2015 recruitment automation, abandoned after it systematically favored male candidates because past successful hires were mostly men, remains the canonical warning. Virginia's HB 2094, the High-Risk Artificial Intelligence Developer and Deployer Act, draws a regulatory line between black-box systems that autonomously make employment decisions and tools that assist human decision-making. California's No Robo Bosses Act and Massachusetts' FAIR Act, both union-backed 2025 bills, add transparency requirements and prohibitions on algorithmic management. Colorado's AI Act, signed in 2024, imposes documentation obligations on deployers.

The operational model that survives scrutiny is human-agent teaming: AI handles first-pass screening and coordination; recruiters own final decisions, negotiation, and judgment calls. Teams that cut human oversight tend to see quality-of-hire regressions. That matches the shift in technical assessment. Because candidates now use generative AI to write code, platforms evaluate problem-solving process and code quality across 1,000+ skills and 40+ languages; final-output scoring alone is insufficient.

The distributional risk is real. NBER's February 2026 survey of nearly 6,000 executives across four countries found 70% of firms "actively use AI," yet 90% reported no employment or productivity impact over the prior three years. Meanwhile, high-frequency payroll analyses show relative employment declines for early-career workers in AI-exposed occupations, while experienced workers hold stable. Brookings flags early distributional risk concentrating at entry-level pathways. For autonomous logistics, where the talent pool is already narrow, over-reliance on automated screens could thin the pipeline further, exactly when the sector needs to widen it.

In a sector where 33 of the top 50 privately held AI companies are based in California alone, the more durable shift, HackerEarth argues, is from "filling seats" to continuous skills verification: assessment data that follows an employee into internal mobility, upskilling, and workforce planning. The startups that treat hiring infrastructure as a compounding asset — not a one-off filter, will be the ones that keep hiring.

The GTM Engineer posting asked for someone to "help Wilson sell himself." The candidates who clear the bot don't just list skills, they hand the agent a spec it can execute.

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.