Ufuk Cem Delice

Independent AI Researcher — UCD Research

About

I am an independent AI researcher focusing on multi-model reasoning, LLM orchestration, and agent-based architectures. My current work targets scalable, reliable, and cost-efficient reasoning systems for production environments.

Research Interests

Multi-Model Reasoning & Dynamic Routing
AI Agents and Orchestration Frameworks
LLM Optimization & Evaluation
Self-Correcting / Debate / ToT methods
High-Load AI Infrastructure

Current Project

Adaptive Routing and Self-Correcting Reasoning in Multi-Model AI Systems
Designing a hybrid reasoning architecture that adaptively routes queries to the most suitable model and applies self-correction to inconsistent reasoning paths.

Benchmarks: GSM8K, HotpotQA, ARC Challenge
Metrics: Accuracy, Consistency, Cost (tokens)
Stack: Python, Docker, LangChain/LlamaIndex
Models: GPT‑4o, Claude 3.5, Mistral Large, Llama 3 (router)

Publications & Preprints

(Upcoming) Adaptive Routing and Self-Correcting Reasoning in Multi-Model AI Systems — to be submitted to IEEE Access. Q4 2025
(Planned) arXiv preprint release. Q4 2025