VSoftware Engineer Blog
Builder · Researcher · LLMs · MLOps · 3D Perception · Switzerland

Vahid Aghajani

I build and research at the edges of applied ML — from LLM systems that reason on messy data, to MLOps pipelines that keep models alive in production, to 3D perception and scene understanding for machines that have to act in the physical world. Seven years turning raw data into shipped ML, across LLMs, computer vision, and production infrastructure. Based in Switzerland.

Connect on LinkedInGitHubYouTubeX

About

I like building end-to-end: from the model that makes the decision, to the API serving it, to the server it runs on. A system isn't finished when the code compiles — it's finished when it's deployed, monitored, and quietly doing its job.

I write on this blog about LLMs, RAG, computer vision, and the practical side of shipping AI — the kinds of notes I wish I'd had while learning them myself.

Tech Stack

Programming Languages

PythonC++JavaScriptTypeScript

Frameworks & Web

FastAPIFlaskReactViteMicroservicesROS2

AI / ML & Computer Vision

PyTorchYOLOOpenCVDetectron2Open3DLLM IntegrationRAG / pgvector

Databases

PostgreSQLPostGISRedisSQLAlchemy

Servers, DevOps & Deployment

DockerDocker ComposeLinux (Ubuntu)NginxGunicornCeleryPrometheus / GrafanaCI/CD (GitHub Actions)HetznerAWSGoogle Cloud

Specialized Domains

GISADASPhotogrammetryPoint CloudsVisual OdometryThree.js

Areas of Expertise

Geospatial & Computer Vision

  • •3D point cloud classification & segmentation at scale
  • •Object detection for infrastructure and photogrammetry pipelines
  • •Sensor fusion for perception systems (LiDAR, camera, depth)

LLM Integrations, RAG & AI Agents

  • •RAG document Q&A with vector search and source grounding
  • •Agentic workflows with tool use and function calling
  • •LLM orchestration across OpenAI, Anthropic, and open-source models
  • •Production ML infrastructure — containerized, monitored, retrainable

Scalable Web Architectures

  • •Full-stack React + FastAPI / Flask platforms
  • •Auth, payments, email, analytics — end-to-end delivery
  • •5+ platforms currently running in production

Servers, DevOps & Deployment

  • •Linux server setup from bare VPS — Docker Compose, Nginx, SSL, backups
  • •CI/CD pipelines and automated deploys via GitHub Actions
  • •Celery workers, cron jobs, Gunicorn — background processing and scheduling
  • •Monitoring and alerting with Prometheus, Grafana, cAdvisor, node-exporter
  • •Operating a fleet of production servers on Hetzner, AWS, and GCP

Let's Connect

I enjoy talking shop with other engineers and researchers — LLM integration patterns, 3D perception pipelines, or just swapping notes on shipping things to production. Always open to a technical exchange.

Connect on LinkedInGitHub