Building scalable web architectures, AI-integrated automations, and secure database systems.
I am a developer focused on Node.js & Python ecosystems. I specialize in modernizing legacy infrastructures, designing relational database schemas (PostgreSQL), and integrating LLMs (OpenAI, Claude) into real-world applications without over-engineering.
- 🔭 I’m currently working on Autonomous AI Agents & SaaS Platforms.
- ⚡ Core Philosophy: "Clean code, scalable database, reliable API."
| Domain | Technologies |
|---|---|
| Backend |    |
| AI Integrations |    |
| Database |     |
| Tools & ORM |     |
Revived a legacy codebase with no documentation and transformed it into a production-ready platform for an early-stage startup.
- Infrastructure Recovery: The production database was missing. I reverse-engineered the data structure solely from Prisma ORM schema definitions and re-initialized a PostgreSQL (Neon) instance.
- CI/CD Pipeline: Replaced manual processes with an automated deployment pipeline connecting GitHub to Vercel, managing environment secrets and domain configurations.
- Cost Optimization: Designed a pre-processing logic that summarizes user inputs before sending them to the AI model, reducing token consumption.
- Security: Engineered an idempotent verification flow and rate limiting to prevent bot spam during user registration.
A multi-tenant AI assistant platform for enterprise clients.
- Fault-Tolerant AI System: Built a 3-layer fallback system that automatically switches providers (Gemini → OpenAI → DeepSeek) in case of API outages to ensure uptime.
- Embeddable Widget: Developed a lightweight
chatbubble.jswidget using Shadow DOM, allowing clients to inject the chatbot into any website without CSS conflicts. - Database Security (RLS): Implemented PostgreSQL Row Level Security (RLS) policies on Supabase to securely isolate User, Partner, and Admin data.
- Document AI (File Search): Integrated OpenAI Assistants API (File Search) to allow users to upload PDF/Docx files, enabling the chatbot to answer questions based on custom knowledge bases.
- Automation: Connected the platform to n8n workflows for cross-platform messaging and task automation (WhatsApp/Instagram).
An end-to-end decision support ecosystem integrating psychometric models with LLMs. Live Demo: kariyermimari.tech
- LLM Integration: Integrated RIASEC and OCEAN models with large language models to generate personalized, data-driven career guidance.
- Data-Driven Logic: Developed a rule-based inference engine on Node.js that transforms raw psychometric survey data into structured role clusters.
- Algorithm Design: Implemented custom algorithms to process user responses and map them to specific behavioral archetypes with high accuracy.