Applied AI Engineer
About sunbeam -----------------
sunbeam is building a fully autonomous SEO agent.
It monitors rankings and competitors, finds content and technical SEO opportunities, writes and improves pages, and publishes to the customer's website.
The agent runs 24/7. It proactively finds work to do, prioritizes tasks, and executes across tools, messaging its manager in Slack when it needs context or approval.
We are not building a chatbot. We are building an employee.
Why us ----------
We ran an SEO agency before building sunbeam. We did SEO for Clay, Replit, Hims & Hers, Rho, and others. We know the work deeply, and we are encoding that expertise into an agent that can do the job with full independence.
We have raised $20M+ from First Round Capital, Basis Set Ventures, WndrCo, and SOMA Capital.
We are starting with SEO, but the long-term goal is every acquisition channel. Marketing services is a $1T/year industry. If AI has automated coding, marketing is next.
The Engineering Challenge -------------------------
Fully autonomous agents are still mostly unsolved.
sunbeam has to plan and prioritize work across weeks and months without a human prompt. It has to reason from messy customer context, durable memory, live search data, CMS state, analytics, and prior decisions. It has to use tools safely, know when to act, know when to ask, recover when integrations fail, and keep improving as it learns more about the company.
The next years of AI will be defined by this shift from "ask → answer" to "observe → act."
What You Will Do ----------------
- Build agent capabilities end to end: planning loops, tool use, prompts and instructions, memory, review workflows, evals, backend services, and customer-facing product surfaces.
- Design safe action systems across integrations like GitHub, Webflow, Google Search Console, Google Analytics, Notion, and Slack.
- Build observability and evaluation workflows so we can understand agent behavior, catch regressions, and measure whether Sunbeam is getting better at the job.
- Debug production agent runs from the model output all the way through database state, tool calls, customer context, and external API behavior.
- Work directly with customers and internal operators to turn expert growth judgment into product behavior.
- Ship quickly across the stack. On a given week, you might change agent instructions, add an MCP tool, create a skill to manage memory, tune a review gate, and fix a frontend workflow.
What We Are Looking For -----------------------
- You have been building AI products for 3+ years, with experience in production LLM agents, evals, tool-use systems, retrieval/memory systems, or workflow automation.
- You are comfortable working across the stack and across levels of abstraction: from "what should the agent do next?" to "why did this Postgres row end up in the wrong state?"
- You have high agency. You notice important problems, form a point of view, and drive them to completion without waiting for a detailed spec.
- You want to understand the customer problem deeply enough to encode judgment into software.
- You are excited by early-stage ambiguity, fast iteration, and a small team where everyone touches product, engineering, design, and customer problems.
Nice to Have ------------
- Experience at a marketing, sales, or growth AI company.
- Experience building from 0 to 1 and 1 to 10 at an early-stage startup.
- Familiarity with C#, .NET, TypeScript, React, GraphQL, Postgres, or MCP-style tool integrations.
The Team --------
Small team, San Francisco office, shipping every day.
- Shravan (CTO), formerly an Engineering Lead at Flixed and a Software Engineer at Facebook, where he worked on mass-scale web scraping for Facebook Jobs and Meta Reality Labs.
- Nico (Product), previously led growth at daydream and ran SEO for Anthropic, Clay, Replit, and others.
- Vishruth (Founding Applied AI Engineer), previously a Data Scientist at Klaviyo specializing in NLP, with a research background from Cornell.
- Daniel (Founding Applied AI Engineer), formerly a Data Engineer on Tesla's Autopilot team and previously Lead Data Engineer at Replit.
- Tom (Founding Designer), who previously led product design at Blockless and now leads our frontend efforts.
Location --------
This role is based in San Francisco and requires in-person collaboration five days a week. We offer paid relocation for candidates outside the Bay Area.
Compensation & Benefits -----------------------
$180,000–$220,000 base salary range plus equity. Medical, dental, and vision insurance. Lunch on in-office days. Wellness and learning stipends. Two retreats a year. Frequent team dinners and outings.
About daydream Labs
Daydream Labs develops a SaaS platform that automates programmatic search engine optimization from end to end.