Software Engineer, Vercel Agent
Vercel
Get hot jobs first on Telegram
New positions appear faster in our channel
- Location
- New York City
- Job Type
- full-time
- Work Format
- ๐ข Hybrid
- Salary
- $232,000 - $348,000
- Posted
- November 20, 2026
Job Description
About Vercel:
Vercel is the agentic infrastructure company, freeing people and agents to ship what's next. For more than a decade we've helped builders move from idea to production with speed, security, and exceptional developer experience.
Now we're scaling our products for both agents and people to ship and run software, built in the open and trusted by OpenAI, PayPal, Ramp, Supreme, and millions of developers worldwide.
About The Role:
We’re looking for a Software Engineer to join the Vercel Agent team. Our mission is simple: deploying on Vercel should feel like adding a seasoned infrastructure engineer to your team. The Vercel Agent investigates incidents before you see them, fixes failing builds, reviews code with deep production context, and answers questions about your running app with the authority of someone who has read the logs.
Today, this work spans multiple shipped products: automated code review on every PR, AI-driven investigation of production anomalies, an intelligent support agent that resolves the majority of customer issues without human intervention, and conversational code generation. We’re now unifying these into a single agent that appears pervasively across surfaces (like the Vercel dashboard, Vercel CLI, Slack, and many more), and is capable of multi-step reasoning, code execution in sandboxed environments, and taking real actions on your behalf.
Two things we’re most excited about: self-driving infrastructure, where 90% of what requires manual DevOps today gets handled automatically by the platform; and framework-defined agents, where the agent is customized to each customer’s application, injecting the right skills when they’re needed rather than dumping generic best practices.
This is a small team with an outsized mandate. You’ll work directly with leadership and across product, design, and infrastructure teams to define what self-driving infrastructure looks like — then build it.
If you’re based within a pre-determined commuting distance of one of our offices (SF, NY, London, or Berlin), the role includes in-office anchor days on Monday, Tuesday, and Friday, even if the role is listed as remote. For location-specific details, please connect with our recruiting team.
What You Will Do:
- Design and build agentic systems that autonomously investigate, diagnose, and remediate production issues — from 500 errors and build failures to performance regressions and security anomalies.
- Architect multi-step agent workflows that coordinate across tools, APIs, and sandboxed code execution environments, with durable state that survives crashes and timeouts.
- Build the skills and tool integrations that give the agent deep knowledge of the Vercel platform, such as deployments, functions, logs, metrics, DNS, and the full lifecycle of a web application.
- Ship polished, user-facing experiences across multiple surfaces (Slack, dashboard, CLI) without waiting for a designer to hand you a mockup. This team values engineers with strong product and UI instincts.
- Develop the infrastructure for dynamic skill injection, so the agent acquires new capabilities at runtime rather than through redeployment, routing the right logic to the right context at the right time.
- Collaborate with other engineers on prompt engineering, model evaluation, and retrieval-augmented generation (RAG) pipelines that ground the agent in accurate, up-to-date knowledge.
About You:
- 7+ years of full-time software engineering experience, with strong TypeScript proficiency. Engineers with deep Python experience who are excited to work in TypeScript are also a great fit.
- A track record of openness and adaptability: you’ve shifted into new technologies, new domains, or new roles and thrived. This space is changing faster than anyone can claim expertise; we want people who learn and adapt, not people who play expert.
- Strong product and UI sensibility. You care about the details of how an AI interaction feels to a user, and you can build great experiences without waiting for a designer.
- Full-stack capability with an understanding of infrastructure, CI/CD, observability, and how large-scale web applications operate.
- Experience building or contributing to frameworks, platforms, or reusable abstractions, not just feature work. You know when to build a primitive versus ship a one-off.
- Demonstrated ability to ship independently, influence architecture, and set direction on a small, high-impact team.
- Excellent communication skills, with an ability to translate technical detail into actionable insight for both engineering and partner teams.
Bonus if you:
- Have experience building agentic systems, AI tooling, or automation frameworks, and especially agents in production. You understand the difference between a demo and a system that works reliably at scale.
- Are familiar with LLM APIs (Anthropic, OpenAI, etc.), prompt engineering, or retrieval-augmented generation (RAG).
- Have worked with Vercel, Next.js, or the modern web deployment ecosystem.
- Have experience with sandboxed execution environments, container orchestration, or secure multi-tenant systems.
Compensation & Benefits:
- Competitive compensation package, including equity.
- Inclusive Healthcare Package.
- Learn and Grow - we provide mentorship and send you to events that help you build your network and skills.
- Flexible Time Off.
- We will provide you the gear you need to do your role, and a WFH budget for you to outfit your space as needed.
The San Francisco, CA base pay range for this role is $232,000 - $348,000 (Staff level IC and above). Actual salary will be based on job-related skills, level/experience, and location. Compensation outside of San Francisco may be adjusted based on employee location. The total compensation package may include benefits, equity-based compensation, and eligibility for a company bonus or variable pay program depending on the role. Your recruiter can share more details during the hiring process.
Disclosures:
- Privacy: Please review our Job Applicant Privacy Policy for more information on how we handle your data.
- Equal Opportunity: Vercel is committed to fostering and empowering an inclusive community within our organization. We do not discriminate on the basis of race, religion, color, gender expression or identity, sexual orientation, national origin, citizenship, age, marital status, veteran status, disability status, or any other characteristic protected by law. Vercel encourages everyone to apply for our available positions, even if they don't necessarily check every box on the job description.
๐ฏ Who is this job for?
This role is best suited to a Senior or Staff-level software engineer with 7+ years of experience, strong TypeScript or Python skills, and full-stack expertise across infrastructure, CI/CD, observability, and large-scale web applications.
The candidate should be skilled in designing reusable platforms and agentic systems, integrating APIs and tools, managing durable workflows and sandboxed execution, and contributing to LLM, prompt engineering, evaluation, and RAG solutions.
They should be experienced in independently shipping user-facing products across web, CLI, and collaboration platforms, investigating and remediating production issues, influencing architecture, and collaborating effectively with technical and product teams.
๐ฌ Potential Interview Questions
How would you design a durable, multi-step agent workflow that investigates a production incident and survives crashes, retries, and timeouts?
I would model the workflow as persisted state transitions with idempotent steps, explicit checkpoints, and durable task execution. Each tool call would have timeouts, retry policies, correlation IDs, and compensation or approval paths for actions that may change production.
How would you safely allow an agent to execute code or remediation actions in a sandboxed environment?
I would use isolated, short-lived sandboxes with restricted network access, least-privilege credentials, resource quotas, filesystem limits, and strict execution timeouts. Actions affecting customer infrastructure should require scoped permissions, audit logging, policy checks, and human approval when the risk is high.
How would you design tool integrations for deployments, logs, metrics, DNS, and serverless functions?
I would define strongly typed, composable tool interfaces with clear input schemas, authorization boundaries, pagination, rate-limit handling, and normalized error semantics. Tools should expose enough context for reasoning while minimizing sensitive data and supporting observability through structured traces.
How would you reduce hallucinations in an agent answering questions about a running application?
I would ground responses in retrieved, time-relevant sources such as deployment metadata, logs, metrics, documentation, and configuration, while requiring citations or evidence for important claims. The agent should explicitly communicate uncertainty and avoid taking action when required evidence is missing or contradictory.
How would you architect dynamic skill injection so an agent can acquire capabilities at runtime?
I would represent skills as versioned, discoverable packages containing descriptions, tool definitions, policies, prompts, and evaluation metadata. A routing layer would select skills based on the user intent, application context, permissions, and current task state, with validation and rollout controls for newly added skills.
How would you evaluate an agent that investigates and remediates production issues?
I would combine offline benchmark scenarios, replayed incidents, synthetic failures, and online metrics such as diagnosis accuracy, successful remediation rate, time to resolution, escalation quality, and unintended-action rate. Evaluations should test tool use, multi-step recovery, security boundaries, latency, cost, and robustness to incomplete or misleading data.
How would you implement reliable streaming interactions for the agent across a dashboard, CLI, and Slack?
I would use a shared event model for agent runs, with resumable streams, sequence numbers, durable run state, and idempotent event handling. Each surface can render the same underlying events differently while supporting cancellation, progress updates, tool-call visibility, and recovery after disconnection.
How would you handle concurrency when multiple agent runs act on the same deployment or incident?
I would use distributed coordination such as leases, optimistic concurrency control, and per-resource locks where necessary. Every action should verify the latest resource version before execution, remain idempotent, and clearly expose conflicts or stale observations to the agent and user.
What TypeScript abstractions would you create for a reusable agent platform?
I would define typed interfaces for agents, skills, tools, contexts, state stores, policies, and execution environments, with runtime validation at external boundaries. The abstractions should support dependency injection, structured errors, tracing, versioning, and test doubles without hiding important operational behavior.
How would you design observability for an agent operating across LLM calls, tools, and infrastructure APIs?
I would emit correlated traces covering user requests, model calls, retrieval results, tool invocations, state transitions, and final actions, while redacting secrets and sensitive customer data. Dashboards should track latency, token and tool costs, failure modes, retry behavior, safety violations, and outcomes by workflow and model version.
๐ Job Summary
Vercel is an agentic infrastructure company building intelligent tools that help developers and AI agents ship and operate software. As a Software Engineer on the Vercel Agent team, you’ll design multi-step agentic systems, automate incident investigation and remediation, build platform integrations, and create polished experiences across the dashboard, CLI, and Slack. The stack centers on TypeScript, with Python, LLM APIs, RAG, sandboxed execution, CI/CD, observability, and modern web infrastructure. This is a hybrid role in New York City with anchor days Monday, Tuesday, and Friday, offering a San Francisco base salary of $232,000–$348,000 plus equity and benefits—an exciting opportunity to define self-driving infrastructure on a small, high-impact team.
Required Skills
Never miss a JavaScript opportunity
Subscribe to get similar jobs and weekly insights delivered to your inbox
Hiring JavaScript developers?
Post your job to 10,500+ registered developers. Starting free.
See PricingRelated jobs
Is this your listing? Claim or request removal