Moves
Off Drupal, Joomla, Kentico, Umbraco, old WordPress and old PHP before support ends. Inventory first, then a fixed quote.
Professional Makers of Things
We make hard software projects boring. Moves, MVPs, AI and rescues, on the record.
DIGI117 is senior engineers in Vancouver, trained from the metal up, for the projects other teams pass on: old platforms that must move, products that must ship, AI that must earn its cost, and code that must stop breaking. Agents do the typing. We write the spec, review every diff and answer for the result.
How we work, in four lines.
- 10-minute standupsBlockers and decisions only. Status is written, not read aloud.
- Direct answersYou talk to the people doing the work. Reply within one business day.
- Transparent billingQuotes in writing before work starts. Invoices list hours by person, level and task.
- AI with human controlAgents write code and drafts. A person reviews every change before it ships.
What We Make
Five kinds of hard project, plus engineers
We take work with a deadline, a mess, or both. Weekly demos, a shared channel, and decisions written down where you can read them.
MVPs
A first version real users can sign up for, in 8 to 12 weeks. Scope fixed in writing, a demo every week.
Custom Software
Products, internal tools and integrations built to be run and changed, not only demoed.
AI Transitions
AI inside a product or process that already exists. We measure the cost per task before we promise the savings.
Rescues
An app that breaks under real users, often one an AI tool built over a weekend. Five-day audit, then a fixed-price fix.
Engineers
Senior engineers inside your team, any team size. Named people, level on every invoice.
What it costs
Prices on the page, quotes in writing
Moves start at $4,900. Rescue audits are $4,500 and count toward the fix. Most MVPs land between $45,000 and $90,000, after a $4,500 scope sprint. Custom software starts from $25,000. An AI map is $5,900, and a pilot starts from $15,000. All prices in US dollars. Every job is quoted in writing before work starts. (Project work has no hourly rate card. Engineers start from $150 an hour. Every invoice lists hours by person, level and task.)
Free tools
Things we check before we quote. Yours to use.
No signup to use them. They run in your browser and send nothing anywhere.
End-of-support calendar
Drupal, PHP, .NET, Node, Umbraco, Kentico and more: when the security fixes stop, and how many days you have.
AI cost per task
What a task costs by hand, and with a model plus a person checking. Before anyone promises savings.
App health self-check
Ten questions from day one of a rescue, with a to-do list at the end.
Questions for any dev shop
Ten questions that separate a team from a sales deck. Us included.
Since 1994
We have met your legacy system. Some of us met it when it was new.
We have made software since 1994: BASIC on the ZX Spectrum, Assembler on early x86 machines, then every language, database and tool that came after, up to today's AI agents. Oracle and SQL Server from their older versions to the current ones, PHP 4 and TypeScript, CGI scripts and serverless. We shipped on all of it, so moving off any of it holds no mystery.
- 1994
First lines
BASIC on the ZX Spectrum, Assembler on early x86 PCs, MS-DOS and its 640K
- 1990s
Desktop and the early web
C, C++, Pascal, Perl CGI, HTML from before CSS existed
- 2000s
Enterprise
Java, .NET, PHP, Oracle, SQL Server, MySQL, stored procedures and nightly batch jobs
- 2010s
Mobile and cloud
Swift, Kotlin, Python, Node.js, PostgreSQL, MongoDB, AWS, Docker
- Now
Agents
TypeScript, React, Kubernetes, Cloudflare, vector search, AI agents with a person reviewing every diff
~/since-1994
# 1994, ZX Spectrum 10 PRINT "HELLO" 20 GO TO 10 # x86, MS-DOS MOV AH, 09h MOV DX, msg INT 21h # 2026 $ agents run --task "move the site" ✓ templates ported, parity checked ● waiting for a person to review # same habit, 32 years apart: # nothing ships unread.
2026
And the stack this quarter.
The tools change every quarter. The habits do not: write the spec, test before you trust, keep a person on the merge button.
Agents, with a spec
Parallel coding agents work from a written spec and an acceptance list. A senior engineer reviews every diff. Agents never merge their own work.
Evals before release
Every AI feature ships with a test set. A prompt or model change that fails it does not ship, however good the demo looked.
UI that shows its work
Design system first, WCAG 2.2 AA, real phones. Agent screens show what the AI did and wait for a person on the parts that matter.
Backends that hold
TypeScript, Python, Go where it pays. Queues, retries and idempotent jobs, so a timeout does not charge a card twice.
Data you can replay
Postgres first: row-level security, pgvector for search, an audit log on every write, migrations that run forward and back.
Privacy by default
Data stays in your accounts. Model calls go only to vendors named in the proposal, on terms that exclude training on your data, in your region where the vendor offers it. Access logged per person and per agent. How we handle it.
Clouds, as code
AWS, Azure, Google Cloud and Cloudflare, on your accounts. Infrastructure as code, so the next team can read it instead of guessing.
Private AI
When data cannot leave: open-weight models on a dedicated cloud slice, a GPU box on your site, or a desktop unit such as NVIDIA DGX Spark.
Trained models
Small models fine-tuned for one narrow job, when they beat a frontier API on cost, speed or privacy. Measured against the same test set.
Productivity, measured
Hours per feature and cost per task, tracked. Token spend capped per project and billed at cost, on its own invoice line.
Senior, the long way
Plenty of titles say Senior. Ours were earned below the framework.
Our engineers learned the stack from the bottom up, before Stack Overflow could do the remembering. That is why a move, a rescue or an AI build does not stall when the problem turns out to live one layer lower.
Infrastructure
Linux, networks, DNS, CI/CD, cloud and bare metal. We read the logs before we blame the network.
Low level
C, Assembler, memory, threads, profilers. When the framework hides the bug, we go under the framework.
Backends
APIs, queues, caches, transactions that do not lose money. Idempotent by habit, not by incident report.
Data operations
Schema changes, migrations, replication and backups that have actually been restored.
Our Process
Three steps, no surprises
Speed comes from cutting waste, not corners.
Discovery & scope
The first call costs nothing. Then a short first step gives you a written scope, what we would cut and why, and a fixed-cost quote. For a move it is the inventory. For an MVP, AI or a rescue it is a fixed-price step whose fee counts toward the build.
Design & build
Short sprints with a demo every week. AI agents write code and tests; a senior engineer reviews each change. Scope changes are priced before they are built.
Launch & grow
We ship, then support. A clean handoff or an ongoing engineering partner, with no lock-in either way. You own the code.
In a Client's Words
A track record you can check
Numbers in our research link to their sources. Client results are as the client reported them.
Digi117 clearly understood our objectives and stayed on course with that in mind. They have flawless software application development and customer service.
Frequently Asked Questions
What does Digi117 do?
Senior engineers in Vancouver for hard software projects. We move sites and apps off old platforms, build MVPs and custom software, put AI to work in products and processes that already exist, and rescue apps that break. We also add senior engineers to existing teams.
What does Professional Makers of Things mean?
We make working products, not decks. Every engagement ends in something you can run, measure and change, and every decision along the way is written down with its reason.
How do you use AI?
AI agents write code, tests and first drafts on our own tooling. A senior engineer specifies the work, reviews every change and decides what ships. Nothing reaches your product unreviewed. See the Peoplarium estimate for what this changes in hours and cost.
How long does it take to build an MVP?
Most take 8 to 12 weeks from kickoff to launch. We quote after a scoping session, so the timeline rests on the scope, not on optimism.
How much does it cost?
Moves start at $4,900; stores and web apps from $18,900. A rescue audit is $4,500, credited to the fix, and stabilizing is $9,500. An MVP scope sprint is $4,500, and most MVPs land between $45,000 and $90,000. Custom software starts from $25,000. An AI map is $5,900, and a pilot starts from $15,000. Engineers start from $150 an hour. Every job is quoted in writing before work starts. See how we price.
Does Digi117 work with US-based companies?
Yes. Most of our clients are US startups and tech companies. We work in US business hours from Vancouver and communicate in plain English.
What technologies do you use?
Whatever your system runs on. We have written software since 1994, from Assembler and BASIC to Java, .NET, PHP and Oracle, to today's TypeScript, Python, React, Node.js, PostgreSQL, AWS, Google Cloud, Azure and Cloudflare. For AI, OpenAI, Anthropic and custom pipelines. New builds follow the product; moves start from what you have.
Can Digi117 join our existing team?
Yes. Named Digi117 engineers join your team, any team size, on your tools (Jira, Linear, Slack, GitHub) and your process. Most start contributing in the first week.
Where do I start?
Send a few sentences and, if there is code, read access. We reply within one business day with what we would do first and what it costs.
Do you sign NDAs, and who owns the code?
We sign an NDA before any discovery or scoping conversation. You own the code: IP transfers to you in full on final payment.
Tell us what you are building.
We will tell you the fastest route to a working product, what we would cut, and what it will cost. No commitment.
Reply within one business day · No sales sequence, no spam