No recruiters, yet founders read every resume at ten‑person Dedalus Labs.
The Screening Gauntlet
Dedalus Labs moves from application to offer in seven days, per TheAntiJobBoard's data, nearly a third faster than the median B2B infrastructure startup under 50 people, which takes 10 days, as TheAntiJobBoard found. The AI infrastructure company, which builds the compute substrate for an agent-native economy, runs its funnel with no dedicated recruiters on a team of roughly 10, per TheAntiJobBoard's figures. The founders review every resume between running the company.
TheAntiJobBoard's data for similar-stage companies shows 50 to 100 applications arrive within the first two weeks. At Dedalus, a recycled CV gets rejected almost instantly. Ashby's parsing favors applications that mirror the job description's language; candidates who tailor their resume clear that first hurdle at a materially higher rate.
The process runs three stages: an intro call, a technical deep dive, and a final round. No take-home assignment is included. The screen emphasizes relevant experience and autonomous operation over abstract algorithmic puzzles. Culture fit carries equal weight: the careers page explicitly calls for "high agency, independent and kind people who want to change the world."
The final round is the hardest stage, per TheAntiJobBoard. By then, founders have already vetted technical depth. Most roles are based in San Francisco.
One follow-up email at day five roughly doubles reply rates. The median application at this stage gets no response ever.
Eight Open Roles, All On-Site in San Francisco
Dedalus Labs is hiring across eight positions: four full-time engineering roles and four internships. Every position ties directly to Dedalus Machines, the persistent-computer platform that lets developers deploy MCP servers in three clicks and build agents in five lines of code.
Full-Time Engineering Roles
Distributed Systems Engineer: $120K–$250K base, 0.5%–1% equity, bonus eligible. This role owns the control plane that schedules and manages persistent agent compute across the fleet. Candidates need deep experience with consensus protocols, leader election, and state replication at scale; the job description emphasizes building "the fastest persistent computer for AI agents" where every millisecond and syscall matters. Production experience with Rust or C++ in a distributed-systems context is expected, along with familiarity running workloads on bare metal or custom hypervisors.
Systems Engineer: same compensation package. While the Distributed Systems Engineer focuses on the control plane, the Systems Engineer works lower in the stack: virtualization, container runtimes, and the host-level primitives that give agents persistent state. The posting calls for expertise in Linux kernel internals, KVM/QEMU or Firecracker, and storage engines. Candidates who have built custom VMMs or contributed to projects like gVisor, Kata Containers, or cloud-hypervisor will stand out.
Software Engineer, Infrastructure & Performance: $120K–$250K base, relocation assistance and visa sponsorship available. This is the broadest of the senior roles, spanning the full infra stack: networking, observability, autoscaling, and the MCP gateway that unifies model providers and tool servers behind a single API endpoint. The role requires fluency in Go or Rust, experience designing high-throughput APIs, and a track record of profiling and eliminating tail latencies in production systems. Because the gateway sits in the hot path for every agent request, performance regression detection and chaos testing are explicit responsibilities.
Design Engineer: $120K–$230K base, 0.5%–1% equity, bonus eligible. Dedalus treats design as an engineering discipline: this role builds the developer-facing console, the MCP marketplace UI, and the SDK surfaces that let engineers go from idea to deployed agent in minutes. The posting seeks someone who can write production React/TypeScript, design REST and WebSocket APIs, and collaborate with the infra team to expose complex primitives (streaming, auth, model handoffs) through clean abstractions. A portfolio showing developer-tooling or infra-dashboard work is effectively required.
Internship Roles
All four internships pay $7,800–$8,667 per month, per Dedalus Labs' careers page, (roughly $45–$50/hour at 40 hours/week), run on-site in San Francisco, and are structured as seasonal programs with potential conversion to full-time.
Systems Engineer Intern: Works alongside the full-time systems team on virtualization and runtime projects: reducing sandbox startup latency, building snapshot and recovery mechanisms, designing schedulers for concurrent workloads, profiling bottlenecks, improving isolation for multi-tenant execution, building test harnesses that simulate machine loss and degraded networks, and evaluating systems research against production architecture. The posting emphasizes C/Rust and Linux systems programming; prior kernel-module or hypervisor contributions (even in coursework) are a strong signal.
Infrastructure Engineer Intern: Listed on the careers page; detailed project scope not publicly specified.
Design Engineer Intern: Also listed there; details as above.
Product Manager Intern: Also listed there; details as above.
What the Mix Reveals
The ratio — four senior IC roles, zero managers, four internships — signals a team still building its core technology and investing heavily in pipeline. Every full-time role carries equity in the 0.5%–1% range, consistent with a post-seed startup that raised $11M in October 2025, as YouTube's announcement reported, and is converting that capital into headcount (28% six-month growth per Simplify data). The salary bands from Zero G Talent's board data ($170K–$250K median $250K) sit above the posted ranges, suggesting the company has room to compete for candidates who clear its technical bar.
The Skill Blueprint: Technical Requirements That Matter
Dedalus Labs does not hire generalists. The company builds a compute substrate for AI agents — full Linux machines that boot in under 50 milliseconds, persist state indefinitely, and bill only for active compute — and every role on its eight-person engineering slate demands depth in the layers that make that possible. The job postings read like a syllabus for a graduate systems course: virtualization, distributed storage, scheduling, networking, and low-level runtime infrastructure all appear in the same bullet list. Candidates who treat any of those as a black box will not pass the screen.
Languages and Runtime Fluency
Rust is the preferred language across every engineering listing, but the requirement is explicit: "Rust is preferred but not required." What matters is production-grade fluency in a systems-oriented language — C++, Go, C, or Rust — backed by evidence that the candidate has shipped substantial systems software in it. The intern posting mirrors this: the same language preference applies. The signal is not syntax familiarity; it is the ability to reason about memory layout, concurrency primitives, and zero-cost abstractions without runtime surprises.
| Language | Stated Preference | Role Coverage |
|---|---|---|
| Rust | Preferred, not required | All engineering roles |
| C++ | Accepted alternative | All engineering roles |
| Go | Accepted alternative | All engineering roles |
| C | Accepted alternative | All engineering roles |
Systems Fundamentals as Table Stakes
Every posting (full-time and intern) leads with the same fundamentals: those areas. "Strong software engineering fundamentals, including data structures, concurrency, memory, and performance" appears verbatim in the Systems Engineer listing. The intern role copies it nearly word for word. This is not boilerplate. Dedalus Machines runs persistent VMs that must maintain state across migrations, snapshots, and scheduler decisions; a candidate who cannot explain how a lock-free queue behaves under contention or why a particular allocator choice affects tail latency will not advance.
"Think abstractions are most useful when you understand what is underneath them." — Dedalus Labs job posting
Domain Depth: Pick One, Own It
The postings list seven deep domains (operating systems, distributed systems, networking, storage, virtualization, runtime infrastructure, and kernel) and expect candidates to demonstrate depth in at least one. The language is precise: "Depth in one or more of operating systems, distributed systems, networking, storage, virtualization, or runtime infrastructure." The intern role adds "Exposure to systems through coursework, research, open source, internships, or self-directed projects" and "Curiosity about how operating systems, networks, storage, and runtimes work beneath their abstractions."
The product roadmap makes the domain priorities concrete:
- Virtualization and isolation: Firecracker, KVM, hypervisor-level sandboxing for secure multi-tenant execution
- Distributed storage: Replication, consensus, consistency models for persistent agent state
- Scheduling: Low-latency allocation across thousands of concurrent agents
- Runtime infrastructure: Networking and storage across the agent execution path
- Performance tooling: Profiling, debugging, benchmarking across networking, storage, scheduling, and runtime layers
Production Scars Over Classroom Grades
"Evidence of end-to-end ownership in production, open source, research, or technically ambitious independent projects" appears in the full-time posting. The intern version is more specific: "A storage engine, scheduler, database, runtime, compiler, operating system, hypervisor, container system, or distributed service you built" and "A project where you encountered a difficult failure and can explain how you diagnosed it." The first filter is the GitHub profile: "The first thing we look at is your GitHub. Show us things you've built. Personal projects. Research. Hackathons. Operating systems projects. Infrastructure tooling. Homelabs. Open-source contributions."
Debugging and Failure-Mode Reasoning
"Strong debugging skills across systems with many interacting components" and "The ability to reason about performance, reliability, correctness, concurrency, and failure modes" are listed as separate requirements. The product targets — reducing sandbox startup from seconds to milliseconds, designing test infrastructure that simulates machine loss and degraded networks — demand engineers who have chased heisenbugs across kernel, hypervisor, and network boundaries. "Enjoy debugging problems that take days to understand and minutes to fix" appears in the philosophy section of both postings.
Communication as Technical Discipline
"Clear communication and the ability to explain architecture, tradeoffs, and technical decisions" and "Technical writing that clearly explains architecture, failure modes, and engineering tradeoffs" appear in both full-time and intern listings. This is not a soft-skill checkbox. The team is 10 people building infrastructure that other engineers will depend on; a design doc that obscures a consistency-model choice or a postmortem that hides a scheduler race condition creates downstream risk the company cannot absorb.
The Baseline Is Moving
The research shows no legacy stack. Every role touches the same surface area: virtualization, distributed systems, storage, networking, scheduling, orchestration, and runtime. A candidate who has only operated Kubernetes clusters or only written application-level Go services will not meet the bar. The intern pay band ($45–50 per hour) signals that even early-career hires are expected to contribute production-quality code to this surface area within weeks. The skill blueprint is narrow, deep, and non-negotiable.
Beyond the Code: Cultural Fit and Mission Alignment
Dedalus Labs makes its cultural filter explicit on its careers page: "Dedalus is defined by its people. We look for those qualities." That triplet functions as the company's cultural shorthand, and it maps directly to the technical problem they're solving. The mission is to build that substrate, with a flagship product, Dedalus Machines, described as that machine. Full Linux machines in under 50 milliseconds. Persistent runtime. Never sleep. Only pay for active compute. The infrastructure demands engineers who can operate without hand-holding, who default to action, and who treat reliability as a moral obligation rather than a checklist item.
The founder background reinforces this profile. Windsor Nguyen, who co-founded Dedalus with Cathy Di '26 after dropping out of Princeton's Computer Science Ph.D. program, framed the decision in velocity terms: "The only thing I didn't like about a Ph.D. was it takes at least five years, and that is an eternity in AI time." Nguyen's own hiring pattern reveals the cultural template in practice: most of the current staff are fellow Princeton students, recruited through a network built on intrinsic motivation. As Nguyen put it: "It's not like you're just staying up coding on Sunday for the love of the game." That phrase captures the drive the screen selects for. It's not about credentials; it's about the inability to stop building.
Kindred Ventures, which led the $11 million seed round alongside Saga Ventures, articulated the mission alignment from the investor side: "Dedalus Lab's continues its mission of pushing the AI agent landscape forward, providing a scalable platform for agent deployment, and being the backbone for agent developer community." The firm added, "We remain excited about our investment in Dedalus and the Agent-to-Agent future." That language — "backbone," "Agent-to-Agent future" — signals a long-horizon bet on infrastructure that outlives any single model release. Candidates who interview at Dedalus are evaluated against that horizon. The technical assessments test distributed-systems fluency, but the cultural screen tests whether the applicant sees the same future: a world where autonomous agents transact, compute, and persist without human babysitting.
Independence shows up in the product architecture. Persistent runtime means no cold starts, no ephemeral containers that vanish between invocations. Engineers at Dedalus own the full stack from the hypervisor up: kernel tuning, scheduler behavior, memory pressure under sustained agent workloads. That ownership model requires the "high agency" trait: when the p99 latency spikes, the person who pushed the commit is the one who diagnoses it. Kindness, the third pillar, isn't decorative. It's operational. A team debugging a memory leak in a shared persistent fleet needs trust that the colleague on call won't hide information or deflect blame. The screen filters for that trust explicitly.
The Princeton dropout cohort that seeds the team carries a specific cultural signal in Silicon Valley. As Joseph Tso, another Princeton leave-of-absence founder, described it: "Our YC group partners — our mentors — often tell us to wear it with pride: You should present yourselves as dropouts of prestigious universities." Kelvin Yu, founder of NewCo, put it sharper: "No one views you as lower status because you dropped out; if anything, you're viewed as higher status." That signal — choosing velocity over credential — aligns with Dedalus's "high agency" bar. The company doesn't require a dropout narrative, but it hires people who could have dropped out and chosen not to, because the work mattered more.
The compensation band ($170k–$250k base across the four senior engineering roles currently open, plus equity) reflects a market-rate offer for San Francisco infrastructure talent, but the cultural filter is the real differentiator. Dedalus isn't competing on perks; it's competing on the density of high-agency builders per square foot. That density compounds. When the next Y Combinator batch asks for a persistent compute layer that doesn't sleep, the team that ships it is the team that already shares the same mental model of what "never sleep" means at the kernel level. The screen protects that density. It's not a gate; it's a gravity well.
The Application Strategy: How to Stand Out
Dedalus Labs' careers page states it directly: the same cultural filter. That sentence is the filter. Every application decision (resume parsing, screening call, technical assessment) maps back to it. Candidates who treat the process as a checkbox exercise get filtered out before a human reviews their file.
Technical preparation needs to mirror the product. Dedalus Machines is described as that product. That phrase, "persistent computer, agent-native economy," appears on the careers page and in the company's self-description. Candidates who can speak to persistent state management, agent orchestration, or the compute substrate challenges of long-running autonomous workloads demonstrate they've done the reading. The roles list C++, cloud, and distributed systems as core requirements. Your resume and screening answers should show concrete work in those areas, not adjacent ones.
The interview process varies by role and location but follows a consistent spine: application form, screening with the hiring team, technical or local requirements, then feedback. Recruiters and hiring managers "work closely with you and guide you through next steps" and "communicate with you at each step." Use that access. Ask clarifying questions about the technical assessment format. Confirm whether the interview will be virtual or onsite. Most full-time roles are on-site in San Francisco; relocation assistance and visa sponsorship are available for the Software Engineer, Infrastructure & Performance role. If you need either, flag it early.
Cultural assessment is explicit. Prepare behavioral examples that map to the stated values: high agency, independence, kindness. A story about debugging a distributed system failure under time pressure hits "embracing challenges" and "personal accountability." A contribution to an open-source project used at scale hits "innovation at scale" and "achieving excellence together."
The inclusive hiring pledge means the process accommodates different backgrounds, but it doesn't lower the bar. Glassdoor shows 91 interview questions and 94 reviews; candidates talk. The consistent theme: technical depth first, cultural alignment second, no shortcuts.
Final lever: the cover letter. Write a short, specific letter that connects your hardest technical problem to the agent-native compute substrate Dedalus is building. Mention the product by name. Show you understand the mission. That's the application that advances.
Compensation and Perks: What Dedalus Labs Offers
Dedalus Labs posts salary bands that put its San Francisco engineering roles in the top tier of early-stage AI infrastructure hiring. The company's own careers page lists four full-time positions (Distributed Systems Engineer, Systems Engineer, Design Engineer, and Software Engineer, Infrastructure & Performance), each with a base range of $120,000 to $250,000, per Dedalus Labs' careers page, (Design Engineer tops out at $230,000, a figure Dedalus Labs' careers page lists). Zero G Talent's live board data, updated within the past week, shows those same roles now clustered tighter at $170,000 to $250,000 with a median of $250,000, suggesting the company has compressed the band upward as it competes for the same distributed-systems talent pool that OpenAI, Anthropic, and the hyperscalers draw from.
Equity is explicit: 0.5 percent to 1 percent for the three core engineering tracks, plus a performance bonus. The Infrastructure & Performance role does not list an equity percentage on the public page but does call out relocation assistance and visa sponsorship; these are benefits the other three roles also offer, per their LinkedIn postings. Internships pay the same monthly rate across Systems, Design, Infrastructure, and Product Manager tracks, a rate that aligns with top-tier Bay Area programs.
| Role | Base Salary Range (Company Careers Page) | Base Salary Range (Zero G Talent Board) | Equity | Bonus | Visa / Relocation |
|---|---|---|---|---|---|
| Distributed Systems Engineer | $120K–$250K | $170K–$250K | 0.5%–1% | Yes | Yes |
| Systems Engineer | $120K–$250K | $170K–$250K | 0.5%–1% | Yes | Yes |
| Design Engineer | $120K–$230K | $170K–$230K | 0.5%–1% | Yes | Yes |
| Software Engineer, Infrastructure & Performance | $120K–$250K | $180K–$250K | Not listed | Not listed | Yes |
| Intern (all tracks) | $7,800–$8,667/mo | Not listed | — | — | — |
The levels.fyi and Glassdoor aggregates tell a different story (median total compensation of $23,558, Software Engineer at $12,676, Product Manager at $102,028), but those figures blend global data (including India-based roles) and older submissions. For the eight San Francisco roles Dedalus is actively filling, the company's own postings and the first-party board data are the relevant signal.
Beyond cash and equity, the package includes meals and office benefits, and the company states it sponsors visas "for exceptional talents." Referrals double a candidate's chance of reaching the interview stage, per the LinkedIn listings (a practical lever for applicants who know current team members).
The gap between the public global aggregates and the live San Francisco bands is itself a data point: Dedalus prices its AI infrastructure roles at market peak for the geography where it hires, while the broader datasets dilute that signal with lower-cost regions and stale reports. Candidates evaluating an offer should anchor on the $170K–$250K band, the 0.5–1 percent equity grant, and the visa/relocation support, rather than the $23K median that appears in third-party summaries.
The Broader Context: AI Infrastructure Hiring Trends
The AI infrastructure boom is usually described in software terms: model releases, benchmark wars, agentic frameworks. But underneath every API call, every inference request, every LLM-powered workflow is physical infrastructure that has to be built, cooled, powered, and maintained by human hands. The market numbers make the scale undeniable. Fortune Business Insights projects the global AI infrastructure market growing from $75 billion in 2026 to $498 billion by 2034, a 26.6 percent CAGR. Grand View Research puts the 2025 base at $33.5 billion and sees $121 billion by 2033. The AI agents slice alone (Dedalus Labs' stated focus) was estimated at $7.7 billion in 2025 with a 39.5 percent CAGR through 2034.
Budgets are following. Deloitte's December 2025 survey of 515 U.S. leaders at enterprises above $500 million in revenue found 86 percent expect AI infrastructure budgets to increase over the next three years, more than tripling on average, with large enterprises projecting nearly fourfold multiples. Seventy percent already devote at least 10 percent of total IT budgets to AI initiatives. Flexential's 2025 State of AI Infrastructure report shows 90 percent deploying generative AI, 76 percent using GenAI applications such as LLMs, chatbots, and coding assistants, and 42 percent running hybrid cloud for AI workloads. By 2028, IDC expects 75 percent of enterprise AI workloads on hybrid fit-for-purpose infrastructure. Token consumption tells the same story: today 37 percent of organizations consume one to 10 billion tokens monthly and 30 percent exceed 10 billion; by 2028, 61 percent expect to surpass that threshold, roughly doubling in two years. The shift toward reasoning models that generate "thinking tokens" before answering accelerates the curve.
The physical constraints are hardening in parallel. AI data centers need 92 gigawatts of additional electric power by 2027. U.S. data centers drew 4.4 percent of electricity in 2023; projections reach 12 percent by 2028. Rack power densities have doubled to 17 kW. Cooling demands could hit 275 billion liters annually. In many regions the bottleneck is chips or talent; in North America it is increasingly electric power. The grid was not designed for this pace. Procurement timelines for 5 MW+ capacity now exceed 24 months, delayed by skilled labor, supply chains, and utility access. Vacancy rates in key data center markets have plunged to a record-low 1.9 percent.
Talent is the binding constraint. Flexential found 86 percent worried about acquiring or developing specialized talent, 61 percent reporting skills or staffing gaps in managing specialized computing infrastructure, up from 53 percent a year ago. Only 14 percent of leaders say they have the right talent for their AI goals. Fifty-three percent face deficits in data science roles. IDC predicts that through 2027, finding people with the right skills will remain challenging, reducing ROI potential for 60 percent of organizations. A10 Networks reports 18 percent cite lack of skilled personnel as a top obstacle to modernization; at the director level, skills gap ties budget at 26 percent. The IEEE Spectrum forum captured the mood bluntly: billions at play but reluctance to pay even middle-class wages to the engineers who make it work.
Against this backdrop, Dedalus Labs' eight open roles carry a salary band of $170k–$250k (median $250k) per the Zero G Talent board. That band sits at the high end of a market where 58 percent of organizations plan infrastructure modernization within 18 months, 63 percent deploy AI tools with built-in training, and 62 percent run structured in-house programs. The roles demand C++, distributed systems, cloud, and kernel-level fluency, precisely the intersection where token growth, power density, and hybrid architecture converge. Dedalus' multi-stage screen is not gatekeeping for its own sake; it is a filter calibrated to a labor market where the gap between demand and qualified supply is measured in years, not quarters.
Working in AI? Zero G Talent tracks the openings: see every open Dedalus Labs role, browse AI jobs, the companies hiring, and the people building the field.