Software Platform Integration Lead

Job Description

Job Description

Position summary

VADD Construction Services builds and runs mission-critical data center and infrastructure projects as an EPC contractor, owner's representative, engineer, estimator and scheduler. Over the past year we have built a family of custom software tools using AI-assisted development. These tools work. The job now is to turn them from a collection of working prototypes into a production platform that people across VADD actually use.

You will own that platform. You will take over the code as it stands, keep it running, stand up the infrastructure behind it (virtual machines, permissions, backups, feedback loops), integrate the pieces so they talk to each other, and build new tools when a group leader or a field engineer needs one. About 95% of the code you produce will be written with AI agents (Claude Code and similar) rather than by hand; you need to be able to read, debug and correct what the agents produce, and know when to write it yourself. You also need to understand logic and be able to troubleshoot with Claude has issues.

The harder half of the job is people. VADD does not micromanage. Nobody will be ordered to use your tools, and most people have a spreadsheet they trust and a way they have always done things. You will sit with group leaders and entry-level staff, on job sites and in offices, find out what they need, fix what is wrong, and earn adoption one user at a time.

What you will do:

Own production

• Take over the existing VADD tools as they stand: learn the code, the deploy docs, the release notes and the open items for each one.
• Deploy and maintain the hosting: virtual machines (per-project cloud VMs and on-site single-board computers), domains, TLS, access control (zero-trust gateway on the public-facing sites), backups and monitoring.
• Set up users, roles and permissions across the tools; put an internal process in place for issuing and controlling AI API keys.
• Close out pre-release items before anything ships wider (removing hard-coded test keys, gating ungated workflows, finishing release engines that were designed but not built).
• Keep a running log of every tool in development, its status, its owner and its next step.

Integrate

• Build the APIs and shared services (users, badges, projects, jobs) that let stand-alone tools work as modules behind one login.
• Resolve naming drift and duplicate stacks so there is one clear product map.
• Connect tool output to tool input where it saves people work (schedule to estimate, one-line drawing to block templates, field capture to reports).

Build with AI agents (about 95% of your coding)

• Follow VADD's rapid-build process: take a one- or two-page requirements outline from a construction engineer, use AI agents to produce a working alpha, and ship it with a white paper, a user guide, a license review and a built-in feedback channel.
• Read, test and correct what the agents produce. Write by hand when that is faster or safer.
• Use only packages whose licenses allow commercial use without payment; flag anything that would need re-coding if a tool is ever sold.

Drive adoption

• Sit down with group leaders and with entry-level staff, in person and on site, to learn what they do, what is not working and what should be added.
• Run the weekly feature/feedback review and turn what you hear into releases.
• Convince people to move off the spreadsheet they know. Make the tool easier than the old way, show them, and follow up until it sticks.

Report

• Meet with the CEO, COO, and VPs regularly (see Travel) to align on goals and review progress.

The software you will inherit and integrate

VADD has built roughly fifteen custom tools over the past year, about half of them in production use on live projects, the rest field-tested, in alpha, or in the requirements stage. Most share one pattern: a small Python standard-library server, a JSON document store, archive-first file handling, a browser front end, and an AI layer calling a commercial LLM API. One production module runs on a low-code front end with a hosted back end. Several are Android apps or embedded devices. You will not be asked to rewrite any of it on day one; you will be asked to understand it, keep it running and connect it.

By category:

Category

Count

What they do

Typical form

Field operations and workforce

4–5

Workforce analytics and site presence, inspection and punch-list tracking, lockout/tagout planning, on-site location of people and equipment, badge access logging

Web modules behind a common login; handheld and embedded devices; per-job single-board computer

Engineering and design

3–4

Electrical block and rack configuration, redundancy and sizing comparison, single-line drawing with BOM and estimate output

Browser apps on a shared stack; two hosted on public domains

Project controls and finance

2–3

CPM schedule viewing and comparison, capital and cash-flow planning with an AI assistant, project controls and PM/CRM

Web and desktop; one behind an access-controlled domain

Reporting and logging

2

Site-walk capture producing AI-cleaned field reports; daily work logging

Android app; web

Platform and launchers

2

A login/launcher front door for the modules; a desktop launcher for installed tools

Web; Electron

Construction management platform

1

A Procore-class, AI-first CM platform (drawings, submittals, RFIs, scope, risk register) built from the pieces above

Requirements phase; planned per-project cloud VM plus on-site single-board computers

Internal AI tooling

1

A model training and feedback loop drawing on VADD's own data

In development

Deployables. Beyond software, VADD fields physical devices tied to these tools: embedded boards, tablets, QR and badge scanners, cameras, site Wi-Fi, and single-board computers acting as per-job servers. You will coordinate with the embedded-systems group on how these are provisioned and kept current.

Integration targets in the first year. One login and user directory; shared project and job records; workforce and presence data flowing to the field modules; schedule, estimate and drawing tools exchanging data; every tool with a working feedback channel; and the CM platform built from these pieces rather than beside them. You will have the ability to build a team of people to assist internally and externally recruiting or you may act as an individual contributor and have input on selecting someone to manage the team as we grow in size and have more people who need

Required qualifications

Technical

• Bachelor's degree in Computer Science (or a closely related degree with a computer science core).
• Working, detailed knowledge of networking (routing, VLANs, VPNs, DNS, TLS), cyber security (access control, secrets handling, hardening of internet-facing services), databases (relational and document; schema design, backups, migration), and software integration (REST APIs, auth, webhooks, data mapping between systems).
• Can write code by hand in at least one mainstream language (Python preferred; JavaScript/TypeScript, Kotlin or similar welcome) and read code in several.
• Has used AI coding agents (Claude Code or equivalent) to build real software and understands the value of "vibe coding": directing agents, reviewing their output, and knowing when to take the keyboard. This is a 95% AI-assisted role.
• Comfortable deploying and administering Linux servers and virtual machines.

Hands-on background (any one of these)

We are not looking for someone who spent a career at a general contractor. We are looking for someone who has done physical, practical work and understands how a job site or a shop actually runs. Any of the following qualifies:

• Raised on a farm. This is an immediate qualifier.
• Time on a construction site in any role, including high-school, college or internship jobs.
• Work as a mechanic, or in a trade.
• Serious home-improvement or handy experience (you have framed, wired, plumbed or rebuilt something).
• Some exposure to construction or finance in any form: estimating, scheduling, job costing, bookkeeping, lending, or a finance or accounting minor.

People (prerequisite: at least one of these)

A large part of this job is sitting across a table from a superintendent or a first-year engineer and getting them to tell you the truth about a tool, then getting them to use it. We need someone charismatic, relatable and energized by that. One of the following is required:

• Previous employment in sales.
• Competitive sports in high school or college.
• Raised on a farm (meets this criterion as well).

Preferred / strong plus

• Customer-service employment: server, bartender, barista, counter service, retail, call center. Also met by being raised on a farm.
• Android (Kotlin) or embedded (ESP32, Raspberry Pi / Orange Pi, Arduino) experience.
• Experience with low-code platforms (Xano, Adalo or similar) and with migrating off them.
• Exposure to construction scheduling (MS Project, Primavera P6) or estimating.
• Familiarity with Procore or similar construction management software, as a user or a critic.
• Cloudflare, Azure or comparable cloud administration.
• Has shipped something people outside your team used, in any setting.

Travel

About 50% in the first six months, settling to about 25% after that. Travel has three purposes:

• Job sites. Visiting active VADD projects to see how the site runs, how people currently track work, and how the tools are (or are not) being used.
• Group leaders. In-person sessions with VADD group leaders to go through features, gaps and needs.
• CEO reviews. A few days each month for the first six months in Steamboat Springs, CO (or a mutually agreeable location) with the CEO to understand company goals and review progress.

First six months

• Days 1–30. Read every tool's code, README, release notes and open items. Stand up your own dev environment for each. Visit at least two job sites. First CEO session in Steamboat.
• Days 31–90. Take over production hosting and access for every live tool. Close the known pre-release items. Publish the product map (what exists, who owns it, what talks to what). Hold feature sessions with every group leader.
• Days 91–180. Ship the shared login/user/project services and the first cross-tool integrations. Run the weekly feedback review on a steady cadence. Show measurable adoption: named users on named projects, by tool.