From raw sources to AI agents: extraction, pipelines, modeling, warehousing, dashboards, and deployment. One accountable system, zero handoff gaps.
Most data projects break at the handoffs. I build and run the whole path, from the first extraction to the AI agents and automations that act on the results. Nothing gets lost along the way, and your data doesn't just sit waiting to be read.
Infrastructure planning, vendor selection, pipelines, modeling, dashboards, and deployment.
Downed dashboards, broken ETL, lost connections, slow models, and visuals nobody reads.
Spreadsheets, CRM, accounting, ads, project tools, APIs. Cleaned and centralized into one source of truth.
Definitions, priorities, and edge cases documented before the first chart, so the numbers mean one thing.
Deterministic automations where they matter, so agents stop reinventing the wheel on every query.
Alerts, next actions, and automations triggered via API, webhooks, or MCP. Your data stops waiting to be read and starts doing the work.
A law firm with 30+ years of history. 100 to 120 reports and dashboards, with logic in DAX, Power Query (M), and JavaScript, rebuilt as standard SQL on Superset.
Meta, Google, Nextdoor, Mailchimp, and Google Business Profile in one view (Jan 2024 to May 2025): leads, impressions, spend, CTR, conversion rate, and cost per lead, with year-over-year comparisons and close-ratio analysis. Filterable by time, source, and campaign.
Select a screenshot to enlarge.
A large importer of agricultural inputs had an accounting team drowning in manual work. Supplier invoice entry went from 2 to 4 hours a day to 30 minutes at most. Monthly bank reconciliation went from two weeks for two people to 4 hours.
Read the caseFor two years, on demand and reporting to the CTO, I turn ambiguous client requests into clean, automated DOMO reports. One fix cut the virtual-row consumption of shared reports by 80%.
Read the caseSelect a screenshot to enlarge.
View on Upwork ↗ (opens in a new tab)Every review below is quoted verbatim from a closed Upwork contract, attributed to the job title.
“Could not be happier with Dylan. I feel like I found a diamond in the rough. Smart, attentive, technically competent. The deliverable and experience have been fantastic.”“Working with Dylan was a pleasure, he is very independent, very organized... and was able to implement several AI automations to support reporting and sales intelligence activity.”“He is very communicative, and clarifies the path as he works... Highly recommended.”“...brought a thoughtful, user-focused approach to every task. His ability to learn quickly and translate complex data into a cleaner, more intuitive experience made a clear impact on our users.”“Our customer is happy as we were able to deliver something on time and within budget!”“Dylan is amazing! He was very responsive and knowledgeable during the whole process and created a product I was very pleased with.”“Dylan's work is superb! He works fast and takes all feedback into consideration, while also explaining everything and giving tips.”“Fast, detail oriented, and very Knowledgable with Power BI. I would hire again!”“I wanted to take a moment to commend you on the excellent work you've done with the Power BI reports.”“Dylan is easy to work with. He delivered the project on time...”“Great guy to work with when it comes to Zoho analytics.”Scope, milestones or hours, and budget agreed up front. Contract, payments, and history stay protected.
Sources, definitions, business rules, and priorities mapped and signed off before anything is built.
Delivered in increments, with progress you can see at every milestone.
Pipelines, models, and business rules documented, so your team understands and owns what was built.
Pipelines and dashboards watched after launch. When something breaks, changes, or grows, it gets fixed fast by someone who already knows the system.
I own your data infrastructure end to end: source extraction, pipelines, modeling, warehousing, dashboards, deployment, and AI.
When AI is part of the answer, I connect agents to real data and keep the repeatable parts deterministic, so the model doesn't reinvent the wheel on every query.