About
An Alberta engineering company that still does the work.
CollectDev Ltd. was founded in Calgary in 2022 by Rob Runka. The company operates from headquarters in Calgary and Edmonton and works with clients across Canada and the United States. Rob has spent more than fourteen years building production software: leading engineering teams, modernizing systems a business already depends on, and putting AI into environments where a mistake is expensive.
He also owns PomPay Financial, a storefront that takes stablecoins and delivers gift cards and eSIMs, and partially owns and has helped build Stoke, a lender to businesses in regulated markets. They are written up as case studies, not presented as clients. A third product, Tour Time Golf, was sunset in 2024, and its case study says plainly why.
The company is growing toward a broader AI engineering and automation practice. The starting point is the same as the work already done. Find a real operational problem, decide whether software or AI should solve it, build the solution, and stay accountable after it is deployed.
You work with the person responsible for the technical decisions. CollectDev does not present a bench of unnamed consultants, and this site does not invent a larger team than exists.
What CollectDev does
CollectDev is a software development and AI automation company. The work falls into four services: AI and workflow automation, custom software and systems integration, software modernization, and long-term engineering partnership. In practice they overlap. A company in Calgary or Denver that wants an AI assistant to triage customer requests usually also needs the assistant connected to its ERP and CRM, which is integration work, and the system it connects to may need some care before anything is built on top of it.
The clients are established businesses, mostly in Alberta but also across Canada and the United States. They have processes that repeat every day, people who run them, and systems they already pay for. The starting point is a specific operational problem: orders re-keyed between two systems, approvals that wait on a person who is travelling, reports assembled by hand every Monday, documents that someone reads and classifies before any work can start. If AI changes the economics of that problem, we use it. If an integration or a simpler workflow will do the job at lower cost and risk, that is the recommendation, and we say so.
Everything we deliver is built to run in production, inside the systems a business already depends on, with a person accountable for it afterward. That includes the AI work. When agents are involved, they draft and a person approves, autonomy is extended only on measured results, and the controls that matter are enforced in the database rather than in a prompt. The Bridge case study describes how we run our own operation this way.
How Rob works
Rob Runka is the person on the call, on the design, and in the code. His background covers founding engineer roles at lending and payments companies, employment in Alberta's energy sector in trading systems and market data, engineering leadership on consumer platforms, and recent work on frontier-model and agent platforms. The experience page lists it, including the former employers and the product that did not survive.
The habits that carry across all of it: money is stored as integers with a ledger behind it, every external boundary is verified rather than trusted, model outputs are forced through a schema before they are stored, and anything that leaves under a person's name is approved by that person. These are not slogans. They are the reasons the systems in the case studies have stayed in production.
What to expect
A conversation first, on this site or on a free introductory call. Then, if the problem is clear, a scoped proposal. If it is not, an optional paid assessment that you can act on with or without us. The approach page describes the steps.
No package prices are published, because the right scope depends on the operation. No case study on this site names a client without permission, and the numbers in them are measured, not estimated.