HEVAL SÖĞÜT

AI ENGINEER @ LIGHT EAGLE AG, SWITZERLAND

I BUILD THE MACHINERY BEHIND THE MODELS.

Good AI engineering is 20% models, 80% infrastructure — I build the 80%. RAG architectures, LLM-powered pipelines, and intelligent automation systems that survive contact with production.

01

About

Most AI demos die the moment they meet real data. I'm interested in the part that doesn't demo well — ingestion, chunking, retrieval quality, orchestration, evals — the plumbing that turns a model into a system. By day I build RAG knowledge bases and automation pipelines for a Swiss firm; by night I poke at mechanistic interpretability to understand what the other 20% is actually doing.

LOCATION
Türkiye · working remotely with Switzerland
EDUCATION
B.Sc. Computer Engineering, Manisa Celal Bayar University — class of 2027
FOCUS
RAG systems · LLM pipelines · workflow orchestration
CURRENTLY
AI Engineer, Light Eagle AG
02

Experience

  1. 2025 — NOW SWITZERLAND · REMOTE

    AI Engineer / Light Eagle AG

    • Architected a multi-agent orchestration system in LangGraph — orchestrator-worker patterns automating supplier-matching workflows across ingestion, cleansing, and decision stages, coordinated via n8n and Python.
    • Built a production-grade RAG pipeline with ReAct-style dynamic query planning and error recovery, on Pinecone vector embeddings and Supabase (PostgreSQL) with hybrid search tuned for retrieval accuracy.
    • Engineered agentic tool use over MCP, grounding agents to internal REST APIs, database queries, and external data sources with function calling and structured output generation.
    • Developed a BI automation agent ("Portfolio Company Reporter") with full observability — prompt tracing, token-usage monitoring, and error logging — optimized for latency, cost, and data integrity for executive stakeholders.
    • LANGGRAPH
    • MCP
    • RAG
    • n8n
    • PINECONE
    • SUPABASE
  2. EARLIER TÜRKİYE

    Software Engineering Intern / ENM Digital

    Full-stack web development on client projects — React on the front, Node.js services behind it. First taste of shipping code other people depend on.

    • REACT
    • NODE.JS
    • REST
  3. EARLIER USA

    Camp Counselor / Paradise Point Summer Camp

    A summer of leading groups in a second language, on the other side of the planet. Not a tech job — but the best crash course in communication and ownership I've had.

    • LEADERSHIP
    • EN-US
03

Skills

AI / LLM SYSTEMS

  • Claude API & integrations
  • LangGraph
  • LangChain
  • MCP
  • RAG
  • Multi-agent orchestration
  • Agentic reasoning (ReAct · CoT)
  • Prompt engineering

AUTOMATION / DATA

  • n8n orchestration
  • ETL pipelines
  • Pinecone · Supabase · vector DBs
  • API integration
  • Observability & tracing
  • Voice-agent transcription

LANGUAGES

  • Python
  • TypeScript
  • JavaScript
  • SQL
  • Java

FOUNDATIONS

  • Machine learning
  • NLP
  • Data structures & algorithms
04

Projects

05

Contact

Building something where the infrastructure matters?

LET'S TALK hevalsogut@gmail.com ↗