Salary and Benefits
- Fully remote position
- $185k - $225K yearly salary
- Significant equity component
- Fully paid medical, dental, and vision
- Unlimited vacation, 401k, parental leave
- Company laptop
- Yearly onsites
About the Role
Most AI demos never become real products. At Sandgarden, we're past the demo phase; we build AI agents that actually write and evaluate software documentation at production quality. That means teaching agents to judge whether a piece of writing is clear, accurate, and useful, designing retrieval systems that ground output in authoritative sources, and building evaluation frameworks that catch subtle failures before users do.
This is a software engineering role first. You'll design agent architectures, ship production systems, and iterate fast where your decisions shape the product. You won't be fine-tuning models or writing research papers.
We're a small, early-stage company where the distance between an idea and production is measured in days, not quarters.
How We Work
High-velocity: We run fast, experiment constantly, and ship. Lean, autonomous teams, minimal process overhead, maximum iteration speed.
AI-first: We don't just build AI; we use it to build. AI is embedded in how we engineer, how we operate, and how we think about quality across the entire company.
Precise and fast: Quick doesn’t mean sloppy. We care about the details: clear writing, sound architecture, and reliable systems, because our agents have to get the details right, too.
Impact over titles: No task is above or beneath you. We care about what you ship, not where you sit on an org chart.
What you’ll do
Design and ship AI agents that write and evaluate software documentation, not toy demos, but production systems that users depend on.
Own the full agent lifecycle from early prototype through production deployment, including defining the processes and best practices the team follows as the product matures.
Build agent primitives like agentic loops, retrieval and grounding strategies, multi-step orchestration, and memory, then figure out which ones actually work in practice.
Create evaluation frameworks that empirically measure writing quality, consistency, and correctness. If the agent produces bad output, your evals should catch it.
Design RAG pipelines that ground agent output in real, authoritative knowledge sources, not hallucinated confidence.
Shape the roadmap and surface problems nobody's named yet. Prototype capabilities before anyone asks for them. Collaborate with product to decide what to build next.
What you’ll bring
Required:
- Hands-on experience building with LLMs, in production or through substantial personal projects that go beyond tutorials.
- Drive outcomes without perfect direction and find creative solutions around obstacles.
- Evaluate whether written output is actually good: clear, accurate, useful, not just "did the model produce text?"
Things that will make you stand out:
- Strong software engineering fundamentals: systems design, API design, distributed systems, data pipelines.
- Proven experience leading technical projects, e.g. setting direction, raising the bar, and mentoring others.
- Deep familiarity with agent architecture patterns like multi-agent orchestration, eval frameworks, and RAG pipelines.
- Experience as a founder or founding engineer; you know what it's like to build from zero.
About the Company
Sandgarden’s product is called Doc Holiday, an AI-native platform to keep organizations in sync as code ships, starting with release notes and documentation. Two minute demo here.
Sandgarden has raised $9M in funding. We build technology products that sell. We have customers and are generating revenue :)
About the team: We are not assholes. We are collaborative, pragmatic, and outcome-oriented. We are fully remote. We value our lives outside of work and give the space to live them. We prioritize respect, kindness, individuality, and dogs!
