AI Native Developer

Ioachim Lihor

I turn machine-learning research into production software. I specialize in Python and in productizing ML and agentic AI systems, from LangGraph agents and RAG pipelines to containerized services on Kubernetes across AWS, GCP, Azure and on-prem Linux.

ioachim.lihor@gmail.com · +40 767 158 866 · ioachim.com · linkedin.com/in/ioachim-lihor-61b528187 · github.com/ioachim-hub

Contact me LinkedIn GitHub
Portrait of Ioachim Lihor

Featured project

Built, shipped and maintained end to end
news.lihor.ro
News Dashboard Today Feed with ranked technical articles

News Dashboard

Live

An AI research desk for technical news. It brings together the sources developers trust, ranks a personalized Today Feed, writes a daily briefing, and answers follow-up questions with citations.

  • AI layer: LangGraph and LangChain pipelines for briefings, article takeaways, and cited “Ask AI” answers over a pgvector corpus.
  • Ingestion: RSS/Atom, GitHub releases, Hacker News and GitHub trending, and custom scrapers, with source-health monitoring.
  • Everywhere: web app, desktop and Android clients, plus a Google Reader–compatible sync API for third-party RSS readers.
  • Production-grade: Docker, Helm charts, CI/CD with CodeQL and Trivy scans, Codecov, Playwright end-to-end tests, self-hostable.
1,000+commits
v1.161current release
37★on GitHub
MITopen source
PythonFastAPILangGraphPostgreSQL + pgvectorReactTypeScriptHelmDocker
Try it live ↗ Source code Docs

More projects

Master's thesis · NLP

Propaganda Detection with LLMs

Master's thesis in NLP at the University of Bucharest: a benchmark of 12 LLMs (GPT-4o, Claude 3, Mistral, Llama 3) on spotting 14 propaganda techniques in news. It compares zero-shot, few-shot, chain-of-thought and fine-tuning; a fine-tuned GPT-3.5 went from 0.43 to 0.56 micro-F1. The repo also includes a chat app that highlights propaganda in any text.

PythonOpenAI fine-tuningNext.jsVercel AI SDK
View repository → Bachelor's thesis · NLP

Fake News Detection

Bachelor's thesis at the Military Technical Academy: an NLP system and application for detecting fake news. This research became the starting point for my master's thesis on propaganda detection.

NLPMachine learning
View repository →

Experience

4+ years shipping ML to production

AI Native Developer

Nov 2022 – Present
RADCOM · Bucharest
  • Productizing ML projects: turning research into Docker containers and Kubernetes deployments that are highly available and scale.
  • Cloud-native delivery: designing and shipping microservices on AWS, GCP and Azure under strict quality and security standards.
  • Backend engineering: robust REST APIs in Python and FastAPI, with clean code and thorough unit tests.
  • Agent orchestration: stateful, autonomous LangGraph agents that generate code and solve problems in real time, in production.
  • End-to-end PoCs: automating engineering workflows from first logic to deployment through CI/CD, in an Agile cross-functional team.

Machine Learning Engineer

May 2024 – Oct 2025
Nexog Technologies · Bucharest
  • Agentic AI systems: designed and containerized custom ReAct agents with LangChain and LangGraph.
  • Advanced RAG: integrated SQL and relational data into retrieval pipelines for production workloads.
  • Observability: end-to-end monitoring and debugging of high-load LLM pipelines with Kubernetes and LangFuse.

Expertise

Core tools and practices

AI & Agentic Systems

LangGraphLangChainLangFuseRAGReActLLM APIsPrompt EngineeringOllamaChromaDBFAISSPyTorchTensorFlow

Productization

DockerKubernetesMicroservicesCI/CDCloud-native Architecture

Backend

Python 3FastAPIPydanticREST APIsSQLUnit Testing

Cloud & Infrastructure

AWSGCPAzureLinux on-premSystem Monitoring

Full-Stack & Data

TypeScriptReactStreamlitGrafanaPower BI

Education

Computer science & NLP research

Master's Degree, Computer Science

2022 – 2024
University of Bucharest

Master's thesis in NLP: propaganda detection with large language models. Deep learning research using PyTorch and TensorFlow.

Bachelor's Degree, Software Engineering

2018 – 2022
Military Technical Academy “Ferdinand I”

Foundations in Python and full-stack development with an emphasis on scalable systems; research on fake news detection.

RomanianNative
EnglishB2 · Independent user

Contact

Open to conversations about ML & agentic AI in production
Emailioachim.lihor@gmail.com Phone+40 767 158 866 LinkedInioachim-lihor GitHubioachim-hub
LocationBucharest, Romania · open to remote
LanguagesRomanian (native) · English (B2)