The organization managed 380+ vendor and contractor relationships across five departments — Procurement, Facilities, HR, Legal, and Administration — through disconnected email chains, spreadsheets, and physical folders. Contract expiration dates were tracked manually in Excel, causing missed renewals and compliance risk. Approval routing was inconsistent and there was no central visibility into SLA compliance or vendor status.
Clinical staff were required to complete monthly supervision documentation for child welfare cases — a regulatory requirement with zero tolerance for missing records. The process required data entry by clinical staff, sequential approvals from Clinical Directors and up to 10 concurrent supervisors, PDF dossier generation, and archival. Everything was done manually via email and physical forms, creating compliance risk and consuming significant staff time.
Monthly facility compliance audits required inspectors to complete detailed checklists across multiple facilities, compile inspection data into formatted PDF reports, route them through management review, and archive approved reports by deadline. The entire process was manual, taking multiple staff-days per cycle and regularly running close to regulatory deadlines.
HR was required to track mandatory certification completion across all staff, assign follow-up training modules based on assessment scores, and maintain documentation for regulatory review. The process relied on manual review of assessment results, manual calculation of required follow-up modules, and manual dispatch of assignment emails — consuming significant HR staff time and creating inconsistencies.
A B2B business needed an integrated, mobile-first inventory and order management system supporting field sales representatives, administrative management, automated transaction processing, real-time inventory updates, and AI-assisted data querying — all within the Microsoft ecosystem without external platforms.
Multi-tool Agentic AI built with Python, LangChain, and Groq. Features job search and CV matching — an autonomous agent that searches job listings and evaluates candidate fit against a CV, leveraging a fast LLM inference backend to deliver real-time recommendations.
End-to-end pipeline architecture designed to automate 800–1,000 daily internal employee requests while maintaining SOX compliance, full audit traceability, and risk-based human oversight. The key architectural decision: rules run before AI — deterministic filtering resolves obvious requests at zero LLM cost; the AI layer only activates for ambiguous cases.
Personal tools and experiments built for learning, exploration, and fun.
Flashcard-style app for learning the 2,000 most frequently used B2 & C1 English words. Features text-to-speech, auto-timer, alphabetical index, level filtering, and bilingual (Turkish + English) display.
Canvas App-style help desk with dashboard analytics, priority tracking, ticket CRUD, comment threads, and status workflows — all client-side with localStorage.
Employee leave request form with manager approval flow, leave balance tracking, and calendar view — simulating a Power Automate approval workflow.
Step-by-step onboarding checklist for new hires, IT, and HR — with progress visualization, task assignment, and completion milestones.