Skyland Performa builds custom systems — document intelligence, AI workflows, and full-stack platforms — that replace the manual work, spreadsheet sprawl, and disconnected tools holding scaling operations back. We come at AI from the data engineering side, which is why our systems survive in production where prototypes fail.
Replace hours of manual review, PDF parsing, and line-by-line data extraction with AI pipelines that turn unstructured documents into clean, structured data in seconds — at the volume your operation actually runs.
Eliminate the manual entry, human routing, and siloed spreadsheets that quietly create data gaps. Replace them with systems where data is captured at the source, validated automatically, and visible in real time across the operation.
Unify revenue, inventory, and communication data trapped in tools that don’t talk to each other. The result is event-driven workflows where data moves the moment it should — no exports, no re-entry, no reconciliation calls on Friday afternoon.
Free your highest-leverage people from administrative friction so they spend their hours on growth, strategy, and judgment calls — the work software can’t replace.
Operations-heavy teams in the $5M to $100M+ revenue range whose workflows have outgrown off-the-shelf tools — typically across professional services, legal, retail and field operations, restaurants and hospitality, and construction. The common thread isn’t the industry; it’s the shape of the problem. High-volume documents. Fragmented data. Manual workarounds that don’t scale. A senior team spending hours on work that software should be doing.
Early-stage startups looking for an MVP or a cheap prototype. Teams shopping primarily on hourly rate. Anyone who wants AI bolted onto a problem it doesn’t actually solve. We build production systems that run core operations — not demos, not chatbots, not proof-of-concept work that gets shelved.
Turns a multi-hour manual insurance review into minutes — recovering $1,500 in a five-minute session, or $5,000 to $8,000+ in 30 to 45 minutes. Based on real Texas claims data.
SupSonic is the SaaS platform I founded and built end-to-end for the roofing industry. Contractors upload a carrier’s insurance estimate. SupSonic identifies what’s missing, underpriced, or has incorrect quantities — generating Xactimate-ready exports with code citations and justification verbiage. In production. Paying customers. Demoing actively across Colorado and Texas.
Saves 7 to 10 hours of back-office work every week. A serverless pipeline processes invoices, deposits, payouts, and credits — including handwritten slips and multi-slip deposit images — scraping the source folder every five minutes and routing data to the right location automatically.
A serverless AP automation pipeline for a multi-location restaurant operation. Documents land in shared Google Workspace folders, trigger an event-driven extraction process, and route structured data into the correct logs by document type. Daily summaries and error notifications route to each location’s inbox every morning.
Google Workspace · JavaScript · Apps Script · Claude API
Ended double-orders and stock-outs across multiple locations by replacing group texts and spreadsheet recounts with a real-time concurrent-ordering system.
A real-time web application replacing manual non-COGS inventory tracking across spreadsheets, group texts, and physical recounts. Solves the concurrent-ordering problem with a two-variable database inventory model: a temporary reservation adjusts immediately when an order is placed, the actual on-shelf quantity updates at pickup. Automated invoicing and notifications fire on each transaction.
Next.js · TypeScript · Python · FastAPI · Supabase
Replaces fragmented intake, scheduling, case management, and client communication tools with a single integrated workflow.
Back-office and front-desk workflow automation for a law firm — intake, scheduling, case management, and client communication consolidated into one system. Currently in build.
Brings data science and predictive modeling to the research-heavy work attorneys bill the most hours on — surfacing the highest-impact case files and precedents first.
A separate engagement focused on the firm’s billable research workload. Early-stage work combining full-stack development with data science, machine learning, and predictive modeling — applied to unstructured case files, historical case performance, and legal precedents. Scope is still taking shape as the firm and Skyland Performa work through what the system should do first.
Next.js · Python · FastAPI · Supabase · AI document extraction · predictive modeling
Recognize your own operation in any of these?
A short call — 30 to 60 minutes — to understand the operational problem, current tooling, and what success looks like. No deck, no pitch. Sometimes the right answer isn’t custom software — an off-the-shelf tool, an integration, or a process change might serve you better. If that’s the case, I’ll point you in that direction. The goal is to solve your problem, not to sell you a build.
For most engagements, the second step is a paid discovery — typically a few hours to a day of focused work to map the actual scope, dependencies, and architecture before quoting a fixed price. This is where most freelance engagements go wrong; doing it properly is non-negotiable.
Weekly check-ins, working software shown early and often. No black-box phases where the client doesn’t see progress for a month. Every engagement runs on a real Git repository with deployment automation from day one — clients get a secure, private preview link where they can test and interact with live updates of their system throughout the build.
Documentation, runbooks, and a transition plan as part of every build. Ongoing maintenance and feature work available month-to-month for clients who want it; clean exit available for clients who don’t.
Most software engineers learned to integrate AI by calling LLM APIs. Skyland Performa came up the other direction: quantitative modeling and data engineering first, full-stack production systems second. When AI shows up in our work, it shows up where it adds value and survives in production.
Skyland Performa was founded by Kurt Baumgardner in 2020. The practice was paused later that year when Kurt accepted the Delivery Director role at the U.S. Small Business Administration during the PPP loan program response, and resumed full operations in 2025.
At SBA, Kurt led IT modernization across the agency’s Office of Capital Access, overseeing platforms that moved over $2B in monthly financial transactions. The flagship engagement was the cloud migration and full redevelopment of the SBA’s loan servicing application — a single platform supporting 2,500+ financial institutions and roughly half of that monthly transaction volume.
Earlier, Kurt co-owned and ran operations at a federal risk analytics firm, where his team built the first lender-level risk measurement and monitoring system for SBA’s 7(a) loan portfolio — quarterly risk reporting across approximately 3,000 lenders, machine learning supporting fraud investigations, and predictive default modeling that fed into the agency’s annual risk plan.
15+ years of hands-on quantitative work — data engineering, modeling, validation strategy — informs how Skyland Performa designs production systems today.
Outside of work, Kurt lives in Denver with his family. He spends his free time snowboarding, hiking the Colorado mountains, catching live music, and following sports — and shares the house with a deeply loved, geriatric family dog who has strong opinions about the daily schedule.
Skyland Performa builds the systems that scaling operations need to move past their current tooling — full-stack platforms, AI automation, document intelligence, and data infrastructure that runs in production.
If that sounds like what your business needs, send a note about the operational bottleneck you’re trying to solve. First conversations are 30 to 60 minutes, no deck, no pitch.
kurt@skylandperforma.comDenver, Colorado · Remote engagements worldwide