About ARC
ARC (The Applied Reports Company) is a multi-tenant applied AI for building engineers.
Building engineers use ARC to deliver the same quality report 50% faster. ARC reads thousands of inputs (photos, notes, drawings, contracts) in a few minutes and shows the engineer relevant context as they write their reports in MS Word. ARC provides generated suggestions, grouped inputs, and deep semantic search across entire projects to take hours out of their workflow.
The product is the ingestion pipeline, meaningful search, and how we work with the engineer. It is NOT a chat app bolted onto CRUD.
The Role
ARC is a small team working in-person, and you'll be one of the senior developers building the core of the product. You'll own big pieces of the system from end to end: the ingestion pipelines that read thousands of inputs per project, the backend resources and GraphQL API on top of them, and the React apps that the users write their reports in. You'll work closely with the CTO and talk directly with the users who use what you build.
Most of the hardest problems span multiple areas: ingestion throughput, search quality, multi-tenant data, and getting LLM output good enough to go into a report an engineer signs. You'll also take your features all the way through to the UI.
We’re a small team (fewer than ten) and we like what we’re building. We work hard, hold each other to a high bar, and pitch in wherever something needs doing. If that sounds like you, we’d like to meet you.
What you'll work on
- Ingestion at scale. Read thousands of photos, scanned PDFs, drawings, and surveys per project in under five minutes, across many tenants at once.
- Search that actually finds things. A search for "roof leak" should turn up the photo of the water-stained ceiling tile, the 2019 roofing permit, and the lease clause on who pays for roof repairs.
- Grouping and evidence. Organize inputs into the systems a report is built around (roof, building envelope, mechanical, and site), with every suggestion linked back to its source.
- Evals and golden data. Grow our handwritten evals and golden datasets so we know a model, prompt, or pipeline change is better before it ships.
- Inference in Elixir. Extend our Elixir-driven numerics and inference for embeddings, ranking, and classification, with no separate Python ML stack.
- Real-time, local-first UX. Keep a local-first React app in sync with long-running jobs over websockets, so users see results stream in as they're ready.
- Working where engineers write. Bring suggestions into our web app and Microsoft Word integration, right in the document the user is editing.
What you'll do
- Architect, design, and build reusable and reliable Elixir applications
- Identify and resolve performance and scalability issues across our pipelines, database, and model calls
- Collaborate across the entire company to build a modern applied AI application
- Own product areas end to end, from talking with customers to shipping and measuring the result
- Drive development through strong product hypotheses: implemented, tested, and shared in the open
- Build the evals and golden data that tell us when an AI feature is ready for customers
- Take features all the way from the backend processing to how the engineer works
What you have
- Deep experience as a back-end or full-stack developer (typically 5+ years), ideally with Elixir, Ruby, or similar languages
- Production experience with data processing pipelines, search, machine-learning systems, and large-scale software systems
- Fast pace and strong craft, with enthusiasm to work in a quick-paced and quality-focused environment
- Analytically rigorous, transparent, and product driven
- Product minded, with the ability to think across the business requirements, the needs of the user, and the technical constraints
- Familiarity with frontend technologies, especially React
Bonus Points
- Experience in an applied AI business
- Experience at a startup, wearing multiple hats
- Experience as an owner of a product area
- Production experience with Elixir, Ash, or Oban
- Experience with retrieval, embeddings, or document understanding (OCR, layout, and vision models)
- Exposure to AEC, real estate, insurance, or other fields where reports carry professional liability