Projects
Real problems. Real solutions. Each project demonstrates how enterprise engineering discipline and AI integration work together in production.
Speech Therapy Report Agent
Audio → AI reports for speech therapists
Speech therapists spend hours writing therapy reports — time taken away from patients. Structured documentation is mandatory but manually time-consuming.
Audio recording of the therapy session → automatic transcription via Whisper (Groq) → structured, formatted AI report in seconds. The therapist corrects instead of writing from scratch.
Fully functional prototype: audio upload → formatted therapy report with Pydantic-validated structure. Demonstrates an end-to-end AI pipeline with real-world application.
RealizeTogether
AI platform for collaborative workflows
Enterprise teams need AI features that are secure, deterministic and integrable into existing system landscapes — not just chatbots.
Production-grade architecture with Next.js frontend, FastAPI backend and Supabase as data layer. AI integration via LangChain with Pydantic Structured Output — no "fire and hope", but validated outputs.
Ongoing R&D platform as testbed for production-grade AI architecture patterns. Authentication, data layer and AI pipeline fully integrated.
This Portfolio
Multi-provider AI backend with fallback resilience
A portfolio backend demonstrating AI capabilities can't fail when a provider hits quota limits or goes down.
FastAPI with LangChain multi-fallback: 4 providers (Google Gemini, OpenAI, Groq, Anthropic) tried in sequence. Pydantic Structured Output for all endpoints — no manual JSON parsing.
CV assistant, vision analysis, sentiment analysis and agentic chat live at sinanucar.com. Resilient operation even during provider outages.