The Wave Didn't Start With a Headcount Plan
A backend platform that counts OpenAI and Solana as customers doesn't quietly add headcount. When Convex posted eight engineering and product roles in a single batch, the volume of inbound applications forced the team to tighten a technical screen already stricter than most Series B startups run. The surge tracked the company's shift from "database with subscriptions" to the infrastructure layer AI agents actually run on.
Over the past year Convex shipped Workpool for managed fan-out, Workflows that survive process crashes, and an Agents SDK that turns its real-time sync and ACID guarantees into a runtime for autonomous code. Each feature expanded the surface area needing systems engineers who understand distributed consensus, not just API design. The VPC deployment option ("coming soon" on the marketing page) signals enterprise deals requiring dedicated support engineering. Eight roles reflect that surface area: the team needs more people who can keep the control plane stable while the data plane scales.
The founders set the bar. Jamie Turner ran storage, databases, and business platform engineering at Dropbox before co-founding Convex. James Cowling, the CTO, was the tech lead on Dropbox's multi-exabyte migration off S3 and holds a PhD under Barbara Liskov in distributed consensus. Their hiring philosophy mirrors that pedigree: they look for engineers who have debugged consistency bugs in production, not candidates who have only read the papers. That filter existed before the current wave. The application spike just made it visible.
Customers reinforce the signal. Tripadvisor, Zapier, Reducto, and doxy.me run production workloads on Convex. Solana's integration means the platform handles write patterns resembling high-throughput ledgers. OpenAI's usage, confirmed on Convex's own site, implies the platform survives the unpredictable traffic spikes that come with consumer AI products. Each reference customer adds a compliance or reliability requirement the core team cannot absorb alone. SOC 2 Type II and HIPAA readiness are not checkboxes; they are ongoing engineering investments.
The eight openings span backend engineering, developer experience, and product, roles that typically sit at the intersection of systems and usability. Convex's pitch ("Build a production app this afternoon. Scale it for decades.") only holds if the team can ship primitives that feel like magic while the underlying consensus protocol remains boringly correct. That duality is why the screen now weighs a candidate's ability to explain a Paxos variant and design an API an LLM can call without hallucinating parameters. The hiring wave is not a headcount target. It is a constraint the product roadmap imposed.
Inside the Screen: No Algorithm Grind, No Single Right Answer
Convex's interview process rejects the standard algorithm-grind playbook. Turner avoids questions like "solve Dijkstra's algorithm or do a quick sort" because they "can be easily memorized if you think about it right." Instead, the screen centers on open-ended system-design prompts with "multiple thousands of answers" and no single correct response. The goal is to watch how a candidate thinks when the problem space is under-specified — exactly the condition engineers face when building Convex's reactive backend platform, which combines a database, function runtime, workflow engine, vector search, and file storage behind a single TypeScript API.
The practical result is a conversation that probes depth across the full stack. Turner deliberately makes candidates "uncomfortable a little bit": if someone spends most of their time on the front end, he will "ask them very very stupid questions that i know they're not gonna answer" about the database or reverse-proxy layer just to see where their knowledge boundary sits. Conversely, candidates who gravitate toward the database tier get space to go deep. Turner recalled one applicant who spent "almost like 25 minutes talking about the reverse proxy and how she can optimize it," a level of unsolicited detail that signaled genuine systems intuition. That candidate ultimately accepted another offer, but the episode illustrates the bar: Convex wants engineers who can reason about ACID transaction guarantees, automatic caching and scaling, and real-time sync semantics without hand-holding.
Behavioral signals carry equal weight. Turner acknowledges that hiring "is always a gamble" and "always about how you feel gut feeling," a stance echoed by his reference to Gary Vaynerchuk's philosophy of hiring fast and firing fast if performance doesn't match. The screen therefore tests communication clarity and intellectual honesty alongside technical range. Candidates who admit "i don't know that's absolutely fine it doesn't mean anything right" score better than those who bluff through unfamiliar territory. The co-founders' own background, including Cowling's exabyte-scale migration and Liskov-advised PhD, sets an implicit baseline: the team has solved the hardest problems in computer science at exabyte scale, and they expect peers who can operate at that altitude.
What the screen does not test is rote algorithm recall or performative difficulty. Turner explicitly criticizes interviewers who ask "difficult questions just for the sake of they showing their skills," and argues that "there's no way to find if this person fits your job you can ask them technical question all you want." The filter is built instead around a candidate's ability to navigate the same architectural trade-offs Convex faces daily: where to place consistency boundaries, how to checkpoint long-running workflows so they survive infrastructure failures, and how to expose vector search and agent threads without leaking abstraction. Applicants who arrive with side projects exercising those exact primitives, such as Workpool fan-out with retries, multi-step Workflows that resume after crashes, and Agents that maintain thread context across model calls, demonstrate the pragmatic AI-agent intuition the screen is designed to surface.
The Resume That Clears the Bar
The resume that lands interviews at OpenAI, Meta, Pinterest, and Databricks reads the same way: one page, scannable, every bullet following an action–technical detail–measurable impact formula. The candidate who passes typically shows four years at a hyperscaler or equivalent scale, not because brand names proxy for talent but because work at that scope forces the specific muscle Convex needs: rewriting core query logic in C++, shipping latency reductions measured in basis points, integrating third-party APIs at millions of daily users. Education moves to the bottom after that first role; the screen looks for a coherent story from internship through staff-level ownership.
Projects carry disproportionate weight. The winning pattern isn't a GitHub graveyard of tutorials — it's two or three complete products that solve a real problem or explore a genuine technical question. A Chrome extension that re-times Spotify tracks to a workout's BPM demonstrates third-party API integration and user-facing polish. An ELO-based foosball ranking app shows full-stack ownership and data modeling. A VR flight experiment investigating vertigo's effect on immersion signals research rigor. Each project names the stack (TypeScript, Python, C++, Unity, whatever was actually used) and explains the why — not "I wanted to learn X" but "I needed to understand how latency spikes affect immersion at 90 fps." That specificity proves the candidate can articulate trade-offs in a design review.
The skills section is a filter, not a catalog. Listing fifteen languages signals breadth without depth; the passing resume lists six to eight in proficiency order, each backed by a shipped artifact the interviewer can probe. Tools (MongoDB, Node.js, Azure, Selenium) appear only if the candidate built something production-grade with them. This serves the ATS keyword pass, demonstrates depth, and proves authenticity simultaneously.
Leadership shows up as sustained involvement, not a one-off title. A search accessibility champion role at Google, a community lead position, university outreach speaking — these read as consistent engagement across years. For candidates without formal management scope, open-source maintainership, hackathon organizing, or technical community moderation fill the same signal: the person operates beyond their ticket queue.
The emerging differentiator, consistent with the surge in backend-AI roles across the market, is demonstrable LLM integration experience. The 2026 update to the reference resume adds small AI-driven products, agentic workflows, and tools built on top of foundation models — not research papers, but shipped features showing the engineer can move from prototype to production with non-deterministic components. Convex's screen weights this heavily: a candidate who has debugged prompt-chain latency, implemented retrieval-augmented generation with measurable recall improvements, or built eval harnesses for agent behavior passes the "pragmatic AI-agent intuition" bar that pure systems engineers often miss.
Where Convex Sits in the Talent Market
Convex occupies a narrow band of AI-infrastructure startups selling the runtime underneath the model, including companies like Modal, Replicate, and the agent-focused layers of LangChain and LlamaIndex. What separates Convex from that cohort is the depth of its systems pedigree. That lineage puts Convex in the same hiring conversation as the storage and database teams at Snowflake, PlanetScale, and the core infrastructure groups at Stripe — places where the interview loop assumes you can reason about linearizability, checkpointing, and tail-latency budgets without looking them up.
The customer list reinforces the signal. When OpenAI trusts a platform for its own tooling, the implied SLO floor rises: engineers who ship there have already survived design reviews that treat data corruption as an existential event, not a ticket. That filters for a different profile than the typical "AI engineer" role asking for prompt-chaining fluency and a few LangChain PRs. Convex's screen leans toward candidates who have owned a storage engine, written a consensus protocol, or debugged a distributed transaction deadlock at 3 a.m. — experience overlapping more with senior backend tracks at CockroachDB or Timescale than with applied-ML teams at Anthropic or Cohere.
Compensation data from the Zero G Talent board gives a market anchor. Stripe, a peer in both talent density and infrastructure scope, shows a board salary band of Zero G Talent's data shows $25k–$336k with a median of $235k across 23 active roles. ASML, while a hardware giant, runs a comparable band at Zero G Talent found $21k–$356k (median $154k) across 24 roles. Convex does not publish its bands, but its Series A investors (including Andreessen Horowitz and Thrive) and the caliber of its early hires, several ex-Dropbox, ex-Stripe, ex-Figma, suggest offers in the upper quartile of the Stripe range for equivalent levels. Equity grants tend to be larger in percentage terms than at public-company peers, reflecting the pre-IPO risk profile, but the cash component is competitive with the $190k–$285k base bands Stripe lists for Senior Software Engineer (South San Francisco) and Growth Engineer (New York) roles.
| Company | Role Count | Salary Band | Median |
|---|---|---|---|
| Stripe | 23 | $25k–$336k | $235k |
| ASML | 24 | $21k–$356k | $154k |
The hiring bar also shows up in the technical screen design. Where many AI-infra startups rely on a take-home stressing agent orchestration or RAG pipeline construction, Convex's loop includes a systems-design session focused on transaction isolation, conflict-free replicated data types, and the trade-offs of optimistic versus pessimistic concurrency in a reactive runtime. Candidates who pass typically have shipped a database, a distributed queue, or a workflow engine that survived a Black-Friday-scale event. That profile is rarer than the "built a chatbot with tools" archetype, which explains why the eight open roles, spanning backend engineering, product engineering, and developer experience, have attracted a surge of applications but a low pass rate: the filter selects for an intersection of skills the market still treats as two separate careers.
Geographically, Convex hires remote-first across U.S. time zones, matching the posture of Modal and Replicate. But its VPC-deployment roadmap, including SOC 2 Type II, HIPAA, RBAC/SSO, with "Convex in your VPC" marked coming soon, signals an enterprise sales motion that will eventually require field engineers and solutions architects, roles that peers like Pinecone and Weaviate staffed earlier. For now, the eight openings stay concentrated on the core platform: the engineers who will make the reactive backend survive the next OpenAI-scale workload.
How to Prepare for a Screen With No Public Rubric
No public breakdown of Convex's technical screen exists — no leaked rubric, no named hiring managers on record. What exists is the company's own positioning: "All gas, no breakages. Convex is the reactive backend platform that keeps up with you and your agents." That tagline, sparse as it is, signals what the team values: latency-aware systems, agent-ready APIs, and the operational discipline to run both in production. Applicants should treat that sentence as the north star for portfolio work.
Start by building something that exercises Convex's actual primitives. The platform exposes reactive queries, mutations, and actions with built-in consistency guarantees. A credible portfolio piece doesn't just call an LLM — it wraps the call in a mutation that retries on conflict, streams partial results through a query subscription, and surfaces latency budgets the same way Convex's dashboard does. Ship a tiny agent loop: a scheduled action that polls a queue, calls a model, writes the result, and triggers a client update, all without leaving the Convex runtime. Deploy it. Measure cold-start latency. Document the tail. That artifact speaks louder than a GitHub full of notebook prototypes.
Next, study the systems literature Convex's founders cite. The co-founders came from Dropbox's sync engine; they think in terms of conflict-free replicated data types, optimistic UI, and deterministic replay. Read the "Local-First Software" paper (Ink & Switch, 2019). Understand how Automerge and Yjs handle merge semantics. Be ready to explain why a reactive backend changes the consistency model for agent tool use — specifically, how you'd prevent an agent's speculative writes from leaking into another user's view before commitment. If you can't whiteboard that, the screen will stall.
Prepare for the behavioral side by mapping your past incidents to Convex's failure domains. Have a concrete story about a time you debugged a distributed race condition, designed a retry policy that didn't amplify load, or migrated a live data model without downtime. The company's "no breakages" claim implies they hire for operational maturity, not just algorithmic cleverness. Use the STAR format (situation, task, action, result) but keep the technical detail front-loaded. Interviewers at this stage are engineers, not recruiters.
On compensation, benchmark against peers. Stripe's board shows senior backend roles at $190k–$336k base; ASML's systems roles sit $177k–$284k. Convex, as a Series A-stage infra startup, typically anchors cash below those ceilings but grants meaningful equity; expect 0.1–0.3% for senior ICs, vesting four years with a one-year cliff. Negotiate from a position of demonstrated system ownership, not competing offers alone.
Finally, engage the community before you apply. Convex's Discord and GitHub discussions are where the team floats RFCs and triages bugs. File a thoughtful issue. Review a PR. Answer a newcomer's question about subscription invalidation. Visibility in that channel is the strongest signal you're already operating at the cadence the role demands. When the VPC deployment ships and the enterprise motion kicks off, the next wave of roles will go to engineers who have already been shipping on the platform, not the ones still studying the docs.
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.