Data, software, and AI systems that hold up in production.
PJK Systems designs and builds the infrastructure small and mid-sized firms run on: data platforms, internal software, and AI tooling. Engagements range from a single build to ongoing technical leadership.
Start a conversationServices
Data engineering
Pipelines, warehouses, and reporting your team can trust.
- Ingestion and orchestration
- Warehouse modeling (dbt)
- Data quality and testing
Software & AI systems
Custom tools built around how your business works.
- Internal applications and APIs
- LLM agents and automation
- Evaluation before rollout
Fractional tech lead
Part-time technical leadership without a full-time hire.
- Architecture and vendor decisions
- Roadmaps and scoping
- Hiring and code review
About
PJK Systems is run by PJ Kelly, a data engineer and team lead based in Columbus, Ohio.
PJ started in healthcare data at Anthem (now Elevance Health), working on financial and project reporting for the core business. At Cureatr, a medication management company, he built and led the data science team, moving manual processes onto automated Python and dbt pipelines. He currently leads data operations at a healthcare AI company.
Healthcare is where most of that work happened: regulated data, legacy systems, and stakeholders who need numbers they can defend. The same approach applies to any business whose operations run on data.
- FocusData platforms, pipelines, ML and AI tooling
- StackSQL, Python, dbt, Dagster, cloud warehouses
- DomainHealthcare: payer, pharmacy, clinical
- EducationMiami University
- Open sourceData engineering toolkit →
How I build
Keep raw data, transform in the warehouse
Business logic changes. When it does, re-run the model, not the extract.
Use the tools before writing tools
Dagster, dlt, and dbt already solve most problems. Custom code that duplicates them drifts.
Tested and fail-closed
Every component ships with tests. Exports refuse to run when tenant isolation can't be guaranteed.
Built for regulated data
Commit guards block secrets and PII before they reach version control.
These patterns are published as an open-source toolkit: github.com/kellypj24/data-engineering
What colleagues say
“PJ not only built and led a team, but he was an advocate for his team, striking a unique balance between forward-thinking strategies and tactical execution.”
“He reliably transforms legacy systems and architecture into modern, performant solutions that precisely meet the needs of his business stakeholders.”
How engagements work
Scope. A short call to understand the problem, then a written proposal with deliverables, timeline, and cost.
Build. Work happens in your environment and repositories, with regular check-ins and working software early.
Hand off. Documentation and tests ship with the code, so your team can run it without me.
Contact
Describe what you're working on and where it's stuck. I reply within one business day.
[email protected]Or connect on LinkedIn.