I turn complicated business problems into software that actually works.

Problem

Internal teams needed AI-assisted workflows for operations, content management, and developer tasks — but the work required deterministic outcomes, not freeform LLM generation. Letting an AI improvise business-critical operations wasn't an option.

Approach

Built custom MCP servers that exposed controlled, deterministic tools to AI systems. The design ensured that operations which needed to be predictable stayed predictable — the LLM handles orchestration, while deterministic tools execute within defined guardrails.

What I Built

Multiple MCP servers providing tool access for content workflows, operational tasks, and developer productivity tooling. Each server defined strict tool interfaces with validation, authorization, and controlled execution paths.

Technical Decisions

Designed the system so deterministic operations remained deterministic. The AI layer handled orchestration and natural-language interaction, while deterministic tools enforced business rules and guardrails.

Outcome

AI-driven workflows for operational, content, and developer tasks with the reliability and predictability of traditional tooling.

Technologies

Problem

Protected digital assets were accessible via direct URLs, allowing unauthorized downloads. The solution needed to enforce access control without disrupting the experience for legitimate users.

Approach

Moved asset delivery behind application logic — every download request was validated against user authentication, authorization rules, and access policies before the asset was served.

What I Built

A secure download system that routed protected assets through the application layer, validating requests against user authentication and authorization rules before serving files.

Technical Decisions

Chose application-level routing over server-config-only approaches to maintain granular control over access policies. Balanced security with usability — authorized users experienced no friction.

Outcome

Reduced unauthorized access without affecting legitimate users.

Technologies

Problem

Bot-driven scraping was causing major bandwidth loss across production web properties.

Approach

Analyzed request patterns, headers, and geographic/IP data to identify scraping traffic and implemented detection and blocking middleware at the application level.

What I Built

Custom detection and blocking middleware using request and header analysis, geographic and IP checks, and rate limiting.

Technical Decisions

Implemented detection and mitigation directly in application middleware, allowing granular inspection of incoming requests and direct enforcement of rate limits and blocking rules.

Outcome

Mitigated bot-driven scraping and reduced abnormal bandwidth consumption while maintaining access for legitimate users.

Technologies

Problem

Content teams were productive in WordPress and didn't want to lose their authoring workflow, but the application needed rendering control, custom business logic, and architectural flexibility that WordPress alone couldn't provide.

Approach

Created an architecture where WordPress remained the content-authoring environment while Laravel controlled application rendering, routing, and delivery. Used a dual-database approach to keep the systems cleanly separated.

What I Built

A dual-platform system with WordPress managing content and Laravel consuming it for rendering. The Laravel layer handled routing, templating, business logic, and performance optimization independently of WordPress.

Technical Decisions

Rather than using WordPress as a headless CMS with a separate API layer, connected Laravel directly to the WordPress database for content reads — reducing complexity and latency while maintaining clean architectural boundaries.

Outcome

Content teams kept their WordPress workflow. The development team gained full control over the rendering layer, application logic, and infrastructure — without the compromises of building everything inside WordPress.

Technologies