About AutoSitu

AI-native workspace for development plan reviews

Tech description:

We’re building multimodal AI systems that reason across thousands of pages of complex architectural and engineering drawings. Our platform combines foundation models, agentic vision, multi-agent workflows, and a knowledge-graph context engine, built with TypeScript, Next.js, Mastra, Postgres, Inngest, and Kubernetes.

Job description:

## Our mission AutoSitu’s mission is to increase global GDP by making development approvals faster, clearer, and more consistent. Every new home, hospital, school, data center, factory, and piece of infrastructure must move through layers of design review, code compliance, coordination, and government approval before it can be built. These processes shape how quickly communities grow, but much of the work is still performed manually across spreadsheets, PDFs, email threads, disconnected regulations, and hundreds—or thousands—of pages of technical drawings. The result is months of avoidable delay, repeated review cycles, costly rework, and experienced professionals spending their time searching for information instead of making decisions. We are building autonomous development-review agents that do the heavy lifting: understanding plans, researching requirements, identifying issues, gathering evidence, and escalating the judgment calls that genuinely require human expertise. If we get this right, housing, infrastructure, and economic development will move faster around the world. ## Who we are AutoSitu (YC W26) is building the AI-native review layer for the built world. Our platform helps cities, architecture firms, developers, and construction teams review complex site and building plans in minutes—catching compliance, coordination, constructability, and QA/QC issues before they become plan-check comments, RFIs, resubmittals, field changes, or months of delay. AutoSitu was founded by George Zhai, whose background spans robotics, autonomy, and applied AI, and Xuanshu “Asher” Lin, who brings deep experience in urban planning, architecture, and development review—combining automation expertise with firsthand knowledge of how regulatory complexity slows real-world projects. We are already supporting public- and private-sector teams on active projects collectively representing hundreds of millions of dollars in development value. We are a small, fast-moving team backed by Y Combinator, building alongside customers responsible for some of the most consequential development decisions in the country. ## What you will own and drive As a Founding Engineer, you will work directly with the founders on difficult applied-AI and distributed-systems problems that sit at the intersection of vision, reasoning, regulation, and the physical world. You will help build systems that can reason across hundreds or thousands of pages of architectural, civil, structural, and engineering documentation—then produce reliable conclusions that expert users can verify. You will: * Build the core intelligence behind AutoSitu’s autonomous review agents * Develop multimodal systems that understand drawings, schedules, details, notes, tables, and cross-sheet references * Design distributed workflows that launch and coordinate specialized agents across many checks and submissions in parallel * Build context and retrieval systems grounded in project documents, local codes, technical standards, and precedent * Create evaluation infrastructure for high-stakes outputs where accuracy, evidence, and judgment all matter * Improve agent reliability through structured reasoning, retries, observability, human feedback, and failure analysis * Own product features end to end—from system architecture and model behavior to UI, deployment, and customer feedback * Work directly with architects, engineers, developers, contractors, and public-sector reviewers to understand how the product performs on real projects * Make foundational decisions about our engineering architecture, product direction, and technical culture ## What to expect in your first 90 days **0–30 days:** Learn the product, customers, review workflows, and core AI architecture. Begin shipping improvements to the agent and evaluation infrastructure. **30–60 days:** Own a meaningful product or intelligence workflow end to end, from technical design through deployment and evaluation on active projects. **60–90 days:** Lead the development and customer rollout of a core capability used in real development-review workflows. You will not spend your first several months waiting for a roadmap or being assigned isolated tickets. You will be expected to understand the problem, make decisions, build the solution, and learn directly from how customers use it. ## We are looking for someone who * Has built and shipped technically ambitious products * Wants to work across AI systems, backend infrastructure, product, and frontend when necessary * Understands that reliability, traceability, and evaluation matter more than impressive demos * Thinks carefully about concurrency, state, retries, idempotency, rate limits, and failure modes * Is deeply curious about frontier multimodal and agentic architectures * Can turn ambiguous, real-world workflows into clear technical systems * Wants to speak directly with customers and understand the domain rather than build from a distance * Moves quickly while maintaining strong engineering judgment * Is comfortable entering a complex industry and learning from experienced practitioners * Wants substantial ownership and responsibility at an early-stage company * Cares about improving how housing, infrastructure, and communities get built Experience in architecture, construction, permitting, or government is not required. A willingness to learn the domain deeply is. ## Our technology We are building distributed, multimodal AI systems that orchestrate specialized agents across hundreds or thousands of pages of architectural and engineering documents. Our platform combines multimodal foundation models, agentic vision, multi-agent workflows, a knowledge-graph context engine, document retrieval, evidence localization, and rigorous evaluation infrastructure. Our stack includes TypeScript, Next.js, Mastra, Postgres, Supabase, Inngest, cloud infrastructure, and modern language and vision models. We care less about whether you have used every tool in our current stack and more about whether you can reason from first principles, choose pragmatic technologies, and build systems that work reliably for real users. ## How we work **Ownership over narrow roles.** Early team members own problems end to end rather than operating within rigid functional boundaries. **Stay close to the customer.** Everyone learns directly from the people using the product. Customer conversations are part of product development, not a separate function. **Evidence over intuition.** We test our systems on real projects, build evaluations, investigate failures, and improve based on what the evidence shows. **Speed with judgment.** We move quickly, but we do not confuse unnecessary complexity with technical sophistication. **Low ego, direct communication.** We challenge ideas openly, acknowledge mistakes quickly, and prioritize the quality of the outcome over who proposed it. **Mission-oriented.** We are building infrastructure for how the physical world gets reviewed and approved. The work is difficult, but the potential economic impact is enormous. ## Why join You will join early enough to shape the product, architecture, culture, and direction of the company. You will work directly with the founders, ship into active projects, interact with customers making real development decisions, and solve applied-AI problems that do not have established playbooks. The opportunity is larger than automating individual plan checks. We are building the intelligence layer through which cities and development teams understand, review, and approve the built world.

Skills:

TypeScript, Distributed Systems

AutoSitu remote jobs

Founding Engineer