AI
Our engineers work with current AI coding agents to modernize legacy code and ETL, ground language models in your own data so every answer can be checked, and design for private models where the data has to stay in your environment.
Our AI services
AI-assisted modernization is where we have client results today. AI on your data and private models are how we approach what comes next. Each starts from the data and systems you already run.
AI-assisted development and modernization
Our engineers work with current AI coding agents on the slow, high-volume parts of modernization: upgrades, remediation, conversion and the tests around them. On one recent engagement that work closed 620+ known vulnerabilities, took Angular 16 to 21 and .NET to version 10, and started consolidating duplicate codebases into one multi-tenant application.
AI on your data
We ground language models in your schema, your business definitions and your access rules, so an answer to a business question comes with the query that produced it and can be verified before anyone acts on it.
AI built into production applications
We build language model features into working applications, from content generation and summarization to extraction from forms and documents, and connect them to the systems and data those applications already depend on.
Private models and AI cost control
Not every workload belongs on a public model API, and not every question needs the largest model. We run private and locally hosted models where data cannot leave your environment, and route each task to the right-sized model so the cost and accuracy of every answer are known before the bill arrives.
AI strategy and roadmapping
We rank AI use cases by business impact and by whether your data can support them, then lay out a phased roadmap.
Our own practice
We build with the same tools we recommend, under written rules about cost and data.
We’re replacing our time-tracking and billing tool with apps built entirely with AI. Testers log bugs in Slack, an AI picks each one up, fixes, tests and redeploys it, and a person tests it before it’s marked done. They roll out across our companies from November.
On our own builds, a small model sorts each task by difficulty and routes it to the model that suits it, so the largest models do only the work that needs them.
Training on your data turned off, your contract’s AI terms always win, including zero data retention, and US-based providers only.
When our engineers work inside your environment, they use the AI tools you provide, and never use company accounts on your data without your permission.
How this connects to AI
Most AI initiatives fail because they start with the technology instead of the data foundation. The AI Enablement Guide covers what the ones that succeed do differently, and how to assess your own readiness.
Our approach
We prove value before scaling investment. Three phases, each with a clear scope and an exit ramp. You decide whether to proceed at each step.
One to two weeks on your data, systems and use cases. It answers three questions: what data do you have, what can it support today and what needs to change before your target use cases become viable.
Deliverable
A readiness scorecard, a 90-day roadmap of use cases ranked by impact and data readiness, a recommended architecture and a scope for the first build.
A proof of value is a working system on your production data that produces a measured business outcome.
Deliverable
A working prototype against production data, measured business outcomes and a production deployment plan.
Ongoing
Production deployment with monitoring, governance and ongoing team support. Managed services, staff augmentation or hybrid delivery based on your team’s needs.
Industries
Patient, pharmacy and clinical data systems held to HIPAA, including where AI is allowed to touch them.
Clients include
ERP, shop-floor and supplier data on one platform, so plant reporting and finance reporting reconcile.
Clients include
Order, inventory, credit and customer data connected across portals, branches and the platforms behind them.
Clients include
Bank data platforms, reporting that holds up to an audit and managed support for the financial systems the business runs on.
Clients include
Claims, member and policy systems for health plans and insurers.
Clients include
Case studies
Retail and distribution
620+
One AI-assisted senior engineer closed every known vulnerability ahead of plan, then started consolidating five to six copies of the codebase into one set of services. About $78K a year projected, over $25K a year already realized.
Retail78%
An AI-assisted tool that converts legacy ETL into tested Databricks pipelines. In a three-month first phase across 177 legacy jobs and scripts, it handled 78% of the legacy transformation types, with tests generated alongside every conversion.
Manufacturing
Three ERPs (IFS, Sage and E2) consolidated into NetSuite and synced with HubSpot, using repeatable cleansing and migration tools built for the client’s data team.
A few practical questions
What buyers usually ask before an AI engagement starts.
Ask us about your situationAdvisory on where AI fits, and the engineering to build it into the systems you already run: assistants and search over your own documents, extraction from forms and invoices, generative features inside existing applications, and AI-assisted modernization of legacy code and ETL. Every build starts from the data it depends on.
Advisory work identifies the AI opportunities worth pursuing, checks whether your data can support them and sets the roadmap. Development is the engineering: integrating models, building the application around them and deploying to production. We do both, so the roadmap is one we know how to deliver.
Yes. We build AI into the platforms clients already run, such as Microsoft Fabric, Databricks, Snowflake, Azure and AWS, and it has to respect the data governance, security and operational workflows already in place.
In phases. The workshop maps the opportunities and assesses data readiness, and the proof of value runs on your real data. Each phase is scoped on its own once we understand the use case and the data behind it, and each ends with a decision about the next.
Related services
Security backlogs, framework upgrades and legacy ETL moved forward by engineers working with Claude and other current models.
Learn moreGoverned, private AI that answers questions from your own data, with the query behind every answer.
Learn moreCustom applications, portals and cloud-native rebuilds of the software that runs the business.
Learn morePipelines, warehousing and real-time reporting built for scale, the foundation AI features depend on.
Learn more
Our own time and invoicing apps were written mostly by AI agents, with a team around them. See how, and what else our engineers build, on SD Labs.
Next step
Tell us the use case you have in mind and the data behind it. Here is what happens after you reach out.
What you want AI to do, the data it would run on and how you would measure it.
Scope comes from your systems as they are, before anything is priced.
Covering scope, team, approach and price.
Angular
.NET Core
Databricks
Claude