I build reliable AI-powered products from interface to infrastructure.

Full-stack engineer shipping UI, reporting, permissions, notification workflows, and production data fixes across the application.

• Implemented application-wide UI updates that aligned complex multifamily workflows with business rules, design standards, and stakeholder expectations.
• Built an Aging Report dashboard from SLA requirements so teams could identify stale cases and prioritize at-risk portfolio work sooner.
• Optimized MariaDB reporting queries behind the dashboard to keep operational views responsive as application volume grew.
• Implemented role-based access control and permission boundaries in Spring Boot to secure data across user tiers and internal workflows.
• Built transactional email templates and notification workflows that improved reliability for stakeholder updates and time-sensitive communication.
• Resolved MariaDB JSON escaping issues that were causing silent production data corruption, protecting data integrity in critical workflows.

Full-stack engineer focused on reporting performance, SLA logic, API reliability, and cloud cost reduction.

• Rebuilt the Aging Report with optimized SQL, indexing, and execution-plan analysis, cutting report generation time on critical operational paths.
• Engineered business-day SLA logic that accounts for weekends and federal holidays so agency reports reflect real compliance windows.
• Built SQL aggregations over live application statuses across disaster-relief programs, giving managers clearer visibility into queue health.
• Replaced AWS Transfer Family with ECS Fargate SFTP, reducing recurring transfer infrastructure costs while preserving secure file workflows.
• Integrated Secrets Manager with singleton connection patterns for RDS Proxy compatibility, improving Lambda reliability under concurrent workloads.
• Hardened API behavior with validation, structured exceptions, and audit-friendly logging to reduce production error rates.

Built the Next.js experience, FastAPI streaming bridge, quiz engine, and agentic document workflow integration.

• Built an AI agent pipeline that parses scanned PDFs into interactive web study interfaces without manual formatting.
• Delivered live progress updates over SSE between FastAPI and Next.js so multi-minute document runs feel transparent instead of blocked.
• Generated structured study guides from raw lecture notes, organizing dense source material into digestible sections and learning flows.
• Built an adaptive quiz engine that generates exam-prep questions from uploaded course material and supports active recall.
• Delivered on-demand concept explanations and deadline prioritization through LLM-powered responses tied to student context.
• Rendered math-heavy outputs with KaTeX so STEM notes remain readable after agentic document extraction.

Designed the plan-execute-observe loop, browser automation layer, responsible scraping controls, and normalized data model.

• Built a Plan-Execute-Observe agentic loop that can discover and extract degree requirements from heterogeneous university catalogs.
• Abstracted extraction logic to support new universities without site-specific hardcoding, reducing manual scraper maintenance.
• Combined OpenAI-powered discovery, SerpAPI, and Playwright to handle both search-driven discovery and JavaScript-rendered catalogs.
• Designed a normalized PostgreSQL schema with async validation for type-safe persistence of courses, prerequisites, and requirement groups.
• Enforced responsible scraping with robots.txt checks, rate limits, crawl delays, and per-domain throttling.
• Tracked token budgets and extraction costs so autonomous catalog discovery stayed predictable and controllable.

Designed conversational requirements gathering, spec generation, task breakdown logic, and memory-aware UX agent behavior.

• Built a full-stack AI project generator that takes users from rough idea to deployable architecture and feature specification.
• Conducted conversational requirements interviews with GPT-4o to turn ambiguous product input into structured JSON specs.
• Generated Cursor-compatible implementation prompts from project specs, accelerating developer execution with clearer task boundaries.
• Built a UI/UX Design Agent with drag-and-drop decision history so design choices remain visible as projects evolve.
• Added duplicate-detection behavior to reduce repeated UX suggestions and keep the agent focused on new decisions.
• Linked product requirements, architecture decisions, and implementation tasks so generated work stays aligned with the agreed plan.

Built agent-powered planning flows, study interactions, shared context, and the interactive visual shell.

• Built planning and study-agent flows that share course, deadline, and user context across product surfaces.
• Designed a real-time Study Buddy experience for explanations, quizzes, and practice without forcing students into a flat chat list.
• Used React Three Fiber and Three.js to create a spatial product shell that makes planning, asking, and practicing feel distinct.
• Applied Framer Motion and glass UI patterns to communicate hierarchy while keeping interactions responsive on desktop and mobile.
• Reduced repeated explanations by keeping agent context available across study and planning workflows.
• Created a more engaging entry point for academic workflows while preserving readable layouts for dense coursework.