Founding Software Engineer - AI & Operations (Full-Stack)

United Concrete; Wallingford (CT); On-site; USD 110,000 - 170,000; Full time

Job summary

United Concrete is seeking an engineer to own the full stack for our production systems at our Connecticut plants. You will drive architecture from data model to API and UI, influencing how we forecast demand and optimize inventory.

You will build OCR + LLM pipelines for invoice auditing, route planning, and predictive maintenance, delivering real business impact within months.

Qualifications

• 3+ years of professional engineering experience, with end-to-end ownership of a production system.
• Strong full-stack fundamentals: backend language, relational design, and a modern frontend framework.
• Experience shipping LLM API integrations into production (OpenAI, Anthropic, etc.).
• Experience with OCR / document extraction and turning scans into usable data.

Responsibilities

• Own architecture, data model, API, and UI for production systems.
• Build OCR + LLM pipeline to audit invoices against pricing.
• Develop demand forecasting and inventory planning using historical data.
• Collaborate with production scheduling, dispatch, and maintenance teams.

Skills

Full-stack development

Production systems ownership

LLM integrations

OCR / document extraction

SQL databases

Frontend framework

Discussions with operations

Tools

OpenAI API

Anthropic API

OCR pipelines

Job description

The short version

We make precast concrete — the stuff underneath roads, buildings, and utilities across the Northeast. Our plants run on spreadsheets, memory, and paper. We’re hiring one engineer to change that, with full ownership of architecture, stack, and roadmap.

You won’t inherit legacy code. You won’t ship features into a backlog nobody reads. You’ll build systems that people 50 feet from your desk use every day — and they’ll tell you, immediately and honestly, whether they work.

Why this role is different

• You choose the stack. Green-field. Data model to UI, it’s yours.

• AI is the job, not a bolt-on. Real problems waiting for it:

  • We recently caught a $6,000 forklift parts invoice that a 5-minute manual price check found for half the cost. We want an OCR + LLM pipeline that catches every one of those automatically.
  • We have 10+ years of sales history and no forecasting. Every busy season, high-demand products stock out. You’ll build the model that flags shortfalls before they happen.
  • A heater once ran all winter and nobody noticed until the bill review — years later. You’ll build anomaly detection on energy and utility data so that never happens again.

• Direct P&L impact. Your code will save real money, measurably, within months — not “improve engagement metrics.”

• You’ll work across the whole business: production scheduling, dispatch and truck routing (including oversized loads), job cost accounting, predictive maintenance, procurement, and compliance.
  What you’ll build (first 12–18 months)

• AI invoice auditing — OCR + LLM pipeline that extracts line items from scanned vendor invoices and checks them against market pricing, with human review where errors are costly

• Demand forecasting — ML/time-series models on a decade of sales data, cross-referenced with mold inventory, to recommend what to pour next

• Unified inventory & job costing — one system for raw materials, finished goods, labor hours, and purchases tied to specific jobs

• Route & load planning — truck routing by weight capacity, trailer selection for over-weight/over-width loads, freight costs flowing into job cost sheets

• Predictive maintenance — evaluate our existing MaintainX deployment (keep it, replace it, or integrate it — your call), then use maintenance history to flag high-wear parts before they fail

• Compliance & alerts — fleet registrations, insurance, inspections, and real-time shipping/receiving notifications

Logistics

• Location: 173 Church Street, Yalesville, CT — onsite with the people you’re building for (adjust if hybrid flexibility is approved)

• Compensation: [$X–$Y base] + [benefits summary] (strongly recommend posting a range — it materially improves reply and apply rates)

• Process: intro call → technical deep-dive on a system you built → short practical exercise (invoice-extraction design) → meet the plant team → offer. Two weeks, start to finish, if you move fast.

• 3+ years of professional engineering experience, including at least one production system you owned end-to-end (data model → API → UI)

• Strong full-stack fundamentals: a modern backend language, solid SQL/relational design, and a modern frontend framework — we care more about judgment than any specific stack

• You’ve shipped LLM API integrations into production (OpenAI, Anthropic, etc.) — structured, task-specific automation, not chatbot demos

• You’ve worked with OCR / document extraction and turned messy scans into clean, queryable data

• You know when a simple regression beats a fancy model — and when AI doesn’t earn its complexity

• You use AI coding assistants (Claude Code, Copilot) daily and ship faster because of it

• You genuinely enjoy sitting down with a dispatcher or plant manager, understanding their manual process, and turning it into software they actually adopt

• You’re self-directed and comfortable being the first engineer

Bonus points: manufacturing/logistics/construction software, route optimization or constraint problems, anomaly detection on time-series data, AI-assisted procurement tools.

You’re probably not a fit if

• You want a large engineering org, a defined career ladder, and someone else deciding the architecture

• You need a product manager between you and your users

• You’re only interested in fully remote work

• Paid Vacation

• Paid Holidays and Sick Time

• Company 401K 10% Match

• Group Medical and Dental