HB hakanbogan.com
senior backend developer,
agentic ai systems
senior backend developer, agentic ai systems

Hakan Boğan

Hakan Boğan, senior backend developer
fig. 01
hakanbogan ~ zsh 40×12 80×24
$
focus/01 Agentic AI workflows
focus/02 RAG and retrieval
focus/03 Distributed backends

profile

I build backend systems, mostly in Python. Django REST and FastAPI services on Kubernetes and AWS, serving thousands of users.

I have led development teams and set how they review and release. The code has gone into logistics fleets, executive vehicles, developer platforms and defence systems. Lately most of my time goes to agentic AI, and specifically to getting several agents to work together and making each one show where its output came from.

skills

skills.toml 46 entries
  • python
  • javascript
  • c++
  • c
  • sql
  • bash
  • django
  • django rest framework
  • fastapi
  • celery
  • kafka
  • rabbitmq
  • redis
  • websockets
  • rest api design
  • grpc
  • llm integration
  • rag pipelines
  • multi-agent frameworks
  • langchain
  • langgraph
  • prompt engineering
  • vector search
  • nlp
  • tensorflow
  • postgresql
  • pgvector
  • mongodb
  • mysql
  • elasticsearch
  • sqlite
  • aws
  • huawei cloud
  • docker
  • kubernetes
  • helm
  • ci/cd
  • mlops
  • argo workflows
  • temporal
  • nginx
  • linux
  • git
  • vue.js
  • html/css
  • qt

selected work

github.com/hakanbogan →
hakanbogan/code-review-agentic-framework Code Review Agents Seven specialised CrewAI agents review a pull request; every finding must point at tool output or a code reference. Python · ★ 5 hakanbogan/self-rag Self-RAG A retrieval-augmented generation system with quality control on its own output. Python hakanbogan/react-agent-calculator ReAct Agent A demonstration of the ReAct (reasoning and acting) agent pattern using LangChain. Python · LangChain hakanbogan/gpt2-turkish-cased GPT-2 Turkish A pretrained Turkish GPT-2, published as an entry point for fine-tuning on other texts. HF model · ★ 16
// icensted 2026 · isbn 978-625-7960-90-8

A Multi-Agent Framework for Evidence-Grounded Automated Code Review

The paper puts a group of LLM agents behind one rule. Nothing reaches the review until it can be traced back to evidence retrieved from the code, so a reviewer can check any comment against what produced it.

→ reference implementation
fig.02 · review pipeline
pull request
retrieved evidence
agents, each with one concern
findings you can check

contact