Selected work / 01
Production operations, applied AI, innovation concepts, and the applications that started my engineering journey.
36 projects
Catch exposed credentials early and guide teams through safe rotation
A developer security concept that detects likely credentials in code changes and configuration, blocks risky merges with evidence, and provides a clear remediation path. It avoids echoing the full secret in alerts or logs.
A production-minded foundation for tenant-aware SaaS products
A platform engineering concept for starting new SaaS products with tenant isolation, authentication, billing boundaries, background work, observability, and usage limits in a coherent foundation.
Understand model spend by product feature and find practical savings
An AI operations concept that attributes language-model usage to product features, customers, prompts, and model choices. It highlights cost and latency patterns and proposes optimization experiments without changing production settings automatically.
Find jurisdiction-specific guidance with source links and effective dates
A source-grounded research assistant concept for businesses checking local permits, licensing, tax, or safety requirements. It asks for the relevant jurisdiction, shows the source and effective date for each answer, and marks unresolved questions clearly.
Map an inherited codebase and plan a phased modernization
A concept for teams inheriting poorly documented repositories. It indexes modules, summarizes business logic, highlights risky files, and proposes a phased migration roadmap that engineers review before acting.
Find untested critical paths and draft the missing tests
A quality concept that scans a repository, ties coverage to user flows, flags untested critical paths, and drafts tests for review. Results appear in CI so regressions in coverage are visible.
Match invoices, receipts, and bank lines, and surface only the exceptions
An agent concept that reads financial documents, matches them to bank transactions, flags duplicates, and prepares an exception report for human approval.
Score candidates against a scorecard with transparent reasons
A hiring concept that parses CVs, compares them against a job scorecard, schedules qualified applicants, and gives recruiters a shortlist with the reasoning visible. Humans make every decision.
Check AI answers against approved sources, claim by claim
A governance concept that extracts claims from an AI answer, checks each against approved documents, and scores support with citations and an audit log.
Some professional work is confidential. Public descriptions stay at a high level and omit customer data, internal systems, and unapproved screenshots.