$250K-$300K Engineering Jobs Now Screen for AI Agent Supervisors

Runtime Bug Detection Changes the Game

Benchmark partner Eric Vishria led a $25 million Series A into Greptile at a $180 million valuation in July 2025, betting that AI code review is moving from static analysis into live execution. The round brings total capital to $30 million and signals a category shift: the review gate is becoming a deployment gate.

Greptile launched in 2023 after its founders — Soohoon Choi, Daksh Gupta, Vaishant Kameswaran, and Davis Liddell — pivoted from an AI shopping assistant called Tabnam during a hackathon. The Georgia Tech graduates entered CREATE-X Startup Launch with the original idea, but customer discovery pushed them toward software development. "First we made the canonical mistake of focusing on attracting investors instead of actual customer needs," Gupta told TechCrunch in June 2024. Y Combinator's Winter 2024 cohort reinforced the lesson.

The product began as an AI code-base expert with an API developers could query directly, not just a chat interface. It helped teams review core changes before merging, find issues, and trace bug origins. By late 2024, over 800 software teams, including Wombo, Metamask, Warp, Exa AI, Bland, and Leya, used it. The company now counts more than 2,000 customers, among them Brex, Whoop, and Substack.

TREX, launched June 15, 2026, runs code and artifacts inside reviews. That moves the review gate from "does this look right" to "does this run right." For engineering managers, the implication is immediate: the hiring profile for the team that owns review tooling shifts from static-analysis specialists to engineers who can supervise agents that execute, observe, and roll back in production-like conditions.

Greptile's own hiring reflects the shift. The company lists 18 open roles on Zero G Talent's board, spanning Staff Generalist Engineer at $250K–300K, Infrastructure Engineer at $220K–300K, Research Engineer at $180K–280K, and Forward Deployed Engineer at $140K–220K, roles built for customers running agents in their own pipelines.

The New Screen: Hiring Engineers Who Supervise AI Code Review

The shift from static analysis to runtime execution changes the hiring screen before the first résumé lands. Greptile's TREX runtime runs code and artifacts inside pull requests, catching bugs static review never sees. That capability forces a different interview question: not "can you write this function?" but "can you scope, constrain, and verify an agent that writes it?"

Greptile's problem statement frames the challenge directly: "Coding standards can be idiosyncratic and are often poorly documented; can we build agents that learn them through osmosis like a new hire might?" The company's Staff Generalist Engineer and Senior Generalist Engineer roles (at $250K–300K and $200K–250K respectively) explicitly target engineers who have operated production systems at Stripe, Google, or Figma and now need to translate that judgment into agent guardrails. The Research Engineer ($180K–280K) and Infrastructure Engineer ($220K–300K) openings reinforce the same profile: people who build the sandbox the agent plays in, not the code the agent produces.

The market signal appears outside Greptile too. CodeRabbit's CEO told TechCrunch that "the unreliability of AI-generated code has given rise to a new corporate role: the vibe code cleanup specialist." That phrase sounds like satire until you see teams budgeting headcount for it. Developers using Copilot and Cursor now spend disproportionate cycles correcting hallucinated logic, missing edge cases, and style drift the model cannot self-correct. Greptile's Forward Deployed Engineer ($140K–220K) and Product Engineer, Security ($170K–300K) roles sit at that intersection: they embed with customers to tune the agent's understanding of idiosyncratic coding standards, which the company's problem statement describes as "poorly documented" standards the agent must learn that way.

That osmosis metaphor is the hiring filter. A senior engineer who can articulate why a pattern matters (not just what the pattern is) becomes the scarce resource. Greptile's Generalist Engineer ($170K–300K) and Security Engineer ($170K–300K) postings emphasize solving hard problems around LLM memory, multi-language codebase indexing, and semantic search for large codebases. The Growth Engineer ($130K–190K) and Frontend Engineer ($160K–250K) roles extend the product surface.

Founders Soohoon Choi, Vaishant Kameswaran, Daksh Gupta, and Davis Liddell built Greptile on the bet that code comprehension — not generation — is the bottleneck. Their Product Manager ($175K–250K) and Founding Product Marketing Manager ($160K–200K) roles now translate that bet into the language of "signal-to-noise ratio" and "PR feedback types customers do and don't care about." The hiring screen has moved: the next senior engineer you hire isn't the one who writes the cleanest PR. It's the one who catches the agent's cleanest-looking bug before it ships.

The 84-Hour Counterpoint: What Candidates Now Screen For

Daksh Gupta posted on X in November 2024 that Greptile "offers no work-life-balance." The tweet described 14-hour workdays, six days a week, sometimes seven (9 a.m. to 11 p.m. or later). It has been viewed more than 1.6 million times and spread across Reddit. Gupta, then 22, framed the schedule as a filter: "The transparency exists to identify" the minority who want this pace. His inbox afterward, he said, split "20% death threats and 80% job applications."

The post did not invent the burnout conversation. It crystallized it. LeadDev's 2025 State of Engineering Management report found 46% of engineers reporting meaningful burnout, 22% critical, 24% moderate. Haystack Analytics puts the number at 83%. LeadDev's Engineering Leadership Report 2025 calls it a crisis driven by layoffs, shifting scopes, and dwindling morale. Velocity up, confidence down.

Gupta's rationale was blunt. "No one cares about the third-best company, or even the second-best company in any category in software," he told Inc. "If you're going to put in 95 percent effort, it's the equivalent of putting in 0 percent effort." He attributed the intensity to San Francisco, not heritage: "I am like this not because I'm Indian but because I'm San Franciscan." He also said the mode isn't sustainable — "It's the first year or two of a startup which is like reaching escape velocity" — and promised adaptation as the company matures and hires people with families.

Candidates noticed. The same engineers who once optimized for equity and learning now screen for sustainability. The applications in Gupta's inbox prove the model still attracts. The death threats prove it repels. Both signals matter. Engineering leaders building agent-supervision teams now face a two-sided screen: they need people who can oversee autonomous code reviewers, and those people are checking whether the overseers get sleep.

The Category Response: Semgrep and Sourcegraph Move to Agentic Code

Greptile's pivot to runtime bug detection with TREX did not happen in a vacuum. Across the developer-tool category, incumbents and best-funded challengers have shipped agentic products that execute changes, not just suggest them. The pattern is consistent: static analysis is being wrapped in planning loops that scope, edit, test, and open pull requests until the code merges.

Sourcegraph made the most visible enterprise move. On July 1, 2026, the company launched Agentic Batch Changes into public beta, an AI agent that scopes, executes, and ships large-scale code migrations across hundreds of repositories until every pull request is mergeable. During the beta, customers merged nearly a thousand changesets created by the agent. The product targets the migration backlog that has historically required weeks of engineer time: framework upgrades, API deprecations, and cross-repo refactors. Sourcegraph's own job board reflects the shift; the company is hiring an ML & Agentic Systems Engineer [IC4] to build the reasoning layer that drives those multi-repo campaigns. Founders Quinn Slack and Beyang Liu have positioned the agent as the answer to "code search at scale" becoming "code change at scale."

Semgrep took a security-first angle. The company introduced nine pre-built Agentic Workflows that hunt subtle authentication, injection, and logic flaws across more than 70 CWEs mapped to the OWASP Top 10. The workflows combine AI reasoning with deterministic program analysis so security teams get critical findings without alert fatigue. Isaac Evans, Semgrep's CEO and co-founder, framed the urgency bluntly: "Attackers are already using AI to find complex exploit chains in minutes." The platform also exposes an SDK so teams can codify custom vulnerability hunts as testable, auditable pipelines.

The competitive set around Greptile is well capitalized. Graphite raised a $52 million Series B led by Accel with participation from Anthropic's Anthology Fund, Menlo Ventures, a16z, and others. CodeRabbit secured a $16 million Series A from CRV. Both sit in the code-review lane Greptile entered, and both move toward agentic workflows that fix, not just flag. The category signal is clear: the hiring screen for the next 18 months will select for engineers who can direct, audit, and harden fleets of coding agents — not just engineers who write clean diffs by hand.

What Operators Should Do Next

The hiring freeze is real. Yale SOM research shows software engineering headcount growth has slowed to 2% annually since ChatGPT's release (down from double-digit rates) while Indeed reports software development postings down 53% from late 2022. Yet starting salaries for CS majors are projected up nearly 7% year-over-year, and 32% of companies expect AI to shrink their workforce by at least 3% within twelve months. The market isn't collapsing; it's bifurcating. Teams that treat AI code review as a productivity multiplier for existing engineers will hire differently than teams chasing headcount reduction.

First, rewrite the req. The Greptile board shows 18 roles added in the past week, none titled "AI code reviewer." The highest-band engineering roles are Staff Generalist Engineer ($250K–300K), Senior Generalist Engineer ($200K–250K), Infrastructure Engineer ($220K–300K), Security Engineer ($170K–300K), and Research Engineer ($180K–280K). The pattern: engineers who can ship product, harden infrastructure, and evaluate model output, not engineers who prompt Copilot. Scale makes this explicit with its Engineering Manager for Agent Oversight role ($248K–310K base), requiring 7+ years engineering experience, 2+ years managing production ML/LLM systems, hands-on familiarity with agent architectures (tool use, planning, multi-agent orchestration), and a track record owning platform-level infrastructure lifecycle. That's the hiring bar for agent supervision.

Second, budget for the evaluation layer. Greptile's TREX runtime bug detector and Sourcegraph's ML & Agentic Systems Engineer [IC4] role signal the same shift: CI/CD now needs observability for agent decisions, not just test coverage. Allocate headcount for engineers who can build evaluation harnesses, regression suites for non-deterministic output, and rollback triggers when agents drift. Don't anchor to 2021 FAANG bands; the premium for agent-oversight experience is already priced in.

Third, treat culture as a filter, not a perk. Founder Daksh Gupta's viral X post (that schedule, "no work-life-balance") didn't just generate outrage; it gave candidates a data point. Yale SOM finds job-market confidence among college grads has fallen from 70% (2022) to 28%, and 19% of degree-holders call it a "good time" to find a quality job versus 35% without degrees. Candidates have leverage to screen employers. If your team runs 84-hour weeks, say so in the job description. If you don't, say that too. The ServiceNow model (CEO Bill McDermott pledging no layoffs, transitioning displaced IT staff into agent-manager roles via internal reskilling) is the competitive counter-narrative. Engineers notice.

Fourth, protect the entry pipeline. Brynjolfsson's Stanford study documents a 16% decline in early-career employment across AI-exposed occupations since late 2022; developer employment ages 22–25 fell nearly 20%. Only 10% of university leaders say graduates are well-prepared for AI-enabled workplaces. If you compress junior roles to fund senior agent-oversight hires, you break your own farm system. Greptile lists a Generalist Engineer role at $170K–300K, a wide band implying they'll take strong new grads and seniors. Structure your ladder the same way: pair each agent-oversight hire with a defined apprenticeship slot, and fund mentorship time explicitly. The alternative is a team that can supervise agents today but has no one to build the next generation of them.

Fifth, watch the category leaders. Semgrep shipped Agentic Workflows for vulnerability hunting; Sourcegraph GA'd Agentic Batch Changes. Both are hiring for agent-fluent engineers. Their job boards (and Greptile's 18 new roles in seven days) are the real-time talent map. Check them weekly. The companies winning the agent-supervision talent war aren't posting "AI engineer" reqs; they're posting platform, infrastructure, security, and research roles with agent evaluation in the requirements. Copy that language.

Benchmark's $25 million bet on Greptile wasn't on a code reviewer. It was on the engineers who will supervise the agents writing the code.

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