Independent examples with test data. They show the implementation approach, not client deployments.
02 / DEVELOPMENT AND REPAIR
Good software can work better.
A missing feature, an awkward import or a bug that keeps returning. I can start with a specific change to the application you already use.
What I can buildA new feature or fix, regression tests and instructions for handing over the change.
An example to exploreAtomic CRM extension
Explore CSV review before writing: a preview, duplicates, invalid rows and explicit contact selection.
Atomic CRM is by Marmelab. My contribution: a contact-import review step. Available to run locally.
03 / INTEGRATION AND AUTOMATION
Stop moving the same data twice.
An order arrives in one system and gets retyped into another. An update disappears after a failure, or arrives twice.
What I can buildA data flow with mapping, error handling and a way to verify what reached its destination.
An example to exploreData Bridge + webhook reliability
Try a controlled flow with retries. Inspect a separate FastAPI and PostgreSQL implementation on GitHub.
04 / MIGRATION AND REPORTING
Change systems. Keep track of your data.
You are moving records, combining exports or explaining different report totals. You need to know what matches and what needs correcting.
What I can buildA validated import, a discrepancy report and a trace back to the source of each issue.
An example to exploreData migration / Revenue BI
Correct a mapping, replay rejected records and compare source with target. Explore where report values come from too.
05 / TEXT PROCESSING AND AI
Less reading. More usable information.
The same information needs extracting from requests and documents. Automation is useful when you can verify the result and catch mistakes.
What I can buildAn extraction workflow with validation, source references and review of uncertain results.
An example to exploreAI workflow lab + change evaluation
Follow a request through fields and sources. Explore a tool that catches regressions before release.
The lab uses synthetic data and simulated actions. The repository demonstrates evaluation; recorded model runs failed acceptance.