- Product-spiked + production agent engineering (harness, evals)
- Cool-startup background, wary of big-tech
- AI-native; can be customer-facing
- Founder mentality, down to grind
Tweets, press, and people that show why this team is worth your time.
Marble is hiring its founding engineer — effectively a second technical co-founder. With 150+ restaurants signed and mass deployment underway, Arjun is the single technical point of failure; he needs someone with immense ownership and agency who can take a big, vague task and drive it to done through long hours and broken things. Engineers here are product builders, not ticket-takers: you'll drive the roadmap, speak with (and get requirements from) customers, sell the product, and have a voice in strategic decisions from day one. There's also a forward-deployed flavor — some restaurants have enough custom flows that this person may end up owning a customer's workflows directly — so strong communication and customer comfort matter.
For this hire specifically, Arjun wants the product-oriented spike (he's deeply experienced on infra and covers it himself) — someone who thinks about users and driving value, with real taste, because Marble "is not a technical product, it has to be tasteful." You'll build the full lifecycle 0→1 (from the agent that drafts a purchase order to the interface a manager approves it in), own the agent harness (context management, orchestration, persistent state, human-approval flows for agents moving real dollars), solve messy data problems (POS/accounting/payroll integrations, pipelines turning invoices/recipes/vendor catalogs/stock counts into clean ground truth), and make agents dependable (evals from real operator workflows, failure tracing, recovery/escalation, model selection on production evidence). Early roadmap includes payments infra (pay vendors directly on Marble) and a proactive "Instinct-for-restaurant-managers" assistant.
What You'll Work On
- Full lifecycle 0→1: customer-facing functionality from the drafting agent to the approval UI
- Own the agent harness: context management, orchestration, persistent state, human-approval flows for purchasing/inventory agents with real dollars attached
- Hard data problems: backend APIs, integrations (POS, accounting, payroll), background jobs, pipelines → clean ground truth
- Make agents dependable: evals from real workflows, failure tracing across model/tool interactions, recovery/escalation, model/latency/cost choices on production evidence
- Drive UX across backend services and React — "working code" → a product operators open every morning
- Learn from production; possible forward-deployed ownership of a customer's custom workflows
Requirements
- A technical spike — for this role, product engineering (React) and/or LLM/agentic applications is the priority (Arjun covers infra); strong fundamentals in Python or JS/TS, willing to become productive in both
- Production agent engineering — shipped agents that use tools and take actions in production; built/debugged the harness (context, tool execution, state, failure recovery, evals); worked across model families/providers
- Evidence of building something substantial they can discuss deeply (architecture, trade-offs, limitations, what they'd change)
- Attention to messy details (missing data, dupes, unexpected inputs, confusing UX)
- LangGraph/LangChain/LangSmith; RAG/retrieval; SQL/NoSQL data modeling; model/tool judgment (agent loop vs. deterministic workflow); evals/tracing/cost/permissions
Execution and Ownership
- Founder mentality — founded something, wants to, or can't stop building; "we'll get it done no matter what," powers through long hours, vague tasks, and things breaking Immense agency and ownership — feels like another technical co-founder; defines the solution rather than implements a spec; argues for the better approach
- Product intuition and taste — cares about the "why," drives user value
- Strong communicator; comfortable talking to customers and gathering requirements directly
Background
- 1–3 years experience (early-career); AI-native
- Cool startup experience strongly preferred — ideally early at something impressive they helped scale (Legora, Rippling-type); wary of pure big-tech (PRD-in, off-at-five mentality — the thing he's filtering against, despite great credentials)
- School used as a signal — a good/known school (Michigan, UIUC, CMU, etc.; not "Stanford or bust")
- Ex-founder / built-something-real signal weighted over GitHub