Engineering
Senior Software Engineer, AI & Agents
Design the tool-loop, prompts, and eval harnesses that turn live customer data into trustworthy, production-grade AI agents.
- TeamEngineering
- LocationRemote (US)
- LevelIC
- Comp$210K – $270K + equity
About the role
We're looking for an experienced engineer to own the applied-AI layer of the product: the agent graphs that plan work, the tool surface that lets them act, and the evaluation harness that keeps quality climbing release over release.
About Lytica Labs
Lytica Labs is building the agentic analytics layer for modern revenue, data, and operations teams. Our platform sits on top of the tools you already use — Snowflake, BigQuery, Salesforce, HubSpot, Stripe, Segment, Slack, Linear, Notion, and a long tail of others — and lets cross-functional teams ask complex business questions in natural language, trigger workflows from the answers, and keep every team aligned on the same source of truth.
Under the hood, we run a multi-model AI stack (OpenAI, Anthropic, and open-source models served via the Vercel AI SDK) orchestrated by a type-safe tool loop that reads live data through Convex and executes side effects through Inngest. The front end is a Next.js 15 App Router monorepo (React 19, TypeScript, Tailwind, Radix) deployed on Vercel; our backend uses Convex for reactive data, WorkOS for auth, Stripe for billing, Resend for transactional email, Sentry for observability, and PostHog for product analytics. Everything is typed end-to-end, everything ships behind feature flags, and everything is observable from day one.
We're a small, senior-heavy team that ships quickly without cutting corners. We believe in clear writing, small PRs, durable abstractions, and treating production traffic like the honor it is. If the intersection of AI agents, real-time systems, and high-leverage analytics is the most interesting problem you can think of right now — we'd love to talk.
What you will do
- Design and ship multi-step agents that reason over customer analytics and trigger downstream actions.
- Build the prompt, tool, and RAG infrastructure that makes those agents reliable across OpenAI, Anthropic, and open-source models.
- Own the evals: offline datasets, online telemetry via PostHog/Sentry, and the feedback loop from customer outcomes back into prompt + model choice.
- Partner with the platform team to expose new data and side-effect primitives as first-class agent tools.
What we are looking for
- 6+ years of software engineering experience with at least 12 months shipping LLM-backed features to real users.
- Fluency with the Vercel AI SDK, tool calling, structured output, and streaming UX patterns.
- Strong instincts for eval-driven development: you trust numbers over vibes.
- You're opinionated about where agents should and shouldn't be used — and can defend those opinions in writing.