Batuhan Yegin
AI Engineer
Five years shipping production software. Now designing AI systems — from RAG architectures and LangGraph agent orchestration to LLM evaluation and fine-tuning. The engineering discipline didn't change. The stack did.
About
The story so far.
Five years, five companies, and a production mindset that ran through all of it. At Uyumsoft I built custom ERP platforms. At Xay Yazılım I shipped four different products across mobile and web. At Confido I developed real-time video communication infrastructure. At NT Bilgi Teknolojileri I architected a full reservation management system. At Supercode, working with Koç Sistem and Yapı Kredi, I built enterprise-grade inventory management systems, SLA engines, SAP integrations, and TÜBİTAK compliance infrastructure. Alongside that, I placed in the top 15 at Teknofest 2024 across two separate categories.
In parallel with all of that, I was studying a different kind of engineering. Not surface-level AI tools — I built my own local LLM engine. I studied agent orchestration patterns, self-correcting RAG architectures, and evaluation frameworks before RAGAS and LLM-as-Judge became standard hiring signals. The production discipline I built in enterprise systems translated directly: reliability, observability, and engineering rigor above all else.
In June 2026, I closed that chapter. I'm now building a public portfolio of 12 AI projects at unickai.com — each one a real, deployed artifact, not a tutorial output.
Experience
- Supercode Koç Sistem · Yapı Kredi2024–2026
- NT Bilgi Teknolojileri Reservation Systems2023–2024
- Confido Real-time Platforms2023
- Xay Yazılım 4 Products2022
- Uyumsoft ERP & Platforms2021–2022
Teknofest 2024 — Top 15, two categories
Available for full-time & freelance
Tech Stack
What I build with.
AI & LLM Ecosystem
LangGraph
Agent Orchestration
RAG / Retrieval
Retrieval-Augmented Gen.
LangChain
LLM Framework
ChromaDB / FAISS
Vector Databases
Pinecone / pgvector
Cloud & SQL Vector DBs
Fine-tuning (LoRA/QLoRA)
PEFT Methods
HuggingFace / PEFT
Model Hub & Fine-tuning
RAGAS
LLM Evaluation
LLM-as-Judge
Evaluation Framework
LangSmith
LLM Observability
MCP
Model Context Protocol
Prompt / Context Eng.
System & Instruction Design
Corrective / Self RAG
Advanced RAG Patterns
Ollama
Local LLM Runtime
OpenAI API
GPT-5.5 / GPT-5.6 / Embeddings
Anthropic API
Claude Opus 4.8 / Fable 5
Whisper + ElevenLabs
Voice AI Pipeline
Guardrails / Safety
LLM Validation & Safety
Engineering
Python
Primary Language
FastAPI
AI API Layer
Pydantic v2
Data Validation & Schema
Docker
Containerization
Celery / Redis
Async Task Queue
PostgreSQL + pgvector
Relational + Vector DB
Nginx
Reverse Proxy & Routing
Linux / Ubuntu
Self-hosted Infra
GitHub Actions
CI/CD Pipelines
pytest / unittest
Test Suite & Coverage
Foundation (5 yrs)
Laravel / PHP
5 yrs · Enterprise production
MySQL / PostgreSQL
Relational DB Design
Redis
Caching, Sessions & Queues
Vue.js 3 / Nuxt
Frontend SPA & SSR
React / TypeScript
Component UI & Typing
Tailwind CSS
Utility-first UI
REST API / GraphQL
API Design & Integration
SAP Integration
Enterprise ERP
Projects
Things I've shipped.
Built at unickai.com — real deployments, not demos.
Building
Docs RAG Assistant
Site/documentation RAG with source citations.
AI CV Analyzer
CV parsing and job-match scoring with structured output.
Self-Correcting Agentic RAG
Corrective RAG + Self-RAG + Adaptive RAG patterns.
MCP Server
Custom Model Context Protocol server and tool integration.
Conversational Support Agent
Multi-turn customer support agent with memory and escalation.
Multi-Agent Research System
Hierarchical multi-agent orchestration for research automation.
LLM Eval Suite
Key SignalRAGAS + LLM-as-Judge evaluation framework for RAG pipelines.
Production AI API
FlagshipFastAPI + LangGraph + MCP + guardrails + observability. Flagship.
Fine-tuned Model App
QLoRA/SFT fine-tuning pipeline with inference API.
Voice AI Assistant
End-to-end voice assistant with Whisper and ElevenLabs.
Contact
Let's build something.
Open to full-time AI Engineer roles, freelance projects, and interesting collaborations. If you're working on something that involves LLMs, agents, or production AI — reach out.