Software Architect
What does a software architect do? I practise architecture as decisions, boundaries and costs: software architecture, system design and event-driven architecture.
What does a software architect do? I practise architecture as decisions, boundaries and costs: software architecture, system design and event-driven architecture.
What does an AI systems engineer do? I design, customize, orchestrate and securely operate AI models as components of a software system; seven areas.
What does a backend developer do? How I work on the API contract, queue, database and server side; since 2010.
What does a DevOps and infrastructure engineer do? Packaging, the CI/CD pipeline, setting up infrastructure in code and shipping to cloud environments.
A PHP developer since 2008 — work shipped on Laravel, Symfony and CodeIgniter, framework-free libraries, and how I actually pick a framework.
What I build in Go — microservices, queue consumers, HTTP APIs and command-line tools, and why I pick Go for that kind of work.
Where I use Python and where I do not — scripting, data wrangling, tooling, and the claims I decline to make, plus the three questions I decide by.
Feeding web and mobile from one API — React Native apps in the stores, contracts carried by TypeScript, and a toolchain that keeps changing.
The work I do with Laravel: APIs, queue-consuming background processes, testing, and when I use what the framework offers and when I leave it.
My work with Docker: packaging applications as containers, Docker as both development environment and deployment unit, alongside Kubernetes and CI/CD.
Running, scaling and restarting containers across several machines with Kubernetes, model serving and inference workloads included.
Describing infrastructure in code and setting it up repeatably with Terraform: cloud resources, the DNS and proxy layer, and the write-plan-apply workflow.
My work with Linux servers: the operating system applications run on, nginx, the firewall, storage, and container and cloud environments.
Model training, fine-tuning and LoRA/PEFT adaptation with PyTorch: where and what for I have used the framework since 2024.
Running training and inference workloads on the GPU with CUDA through PyTorch, TensorFlow and OpenAI Triton: where I have used it since 2024.
Model training with TensorFlow: where I have used the framework since 2024 and how it sits with PyTorch, CUDA and model serving.
Rust for services and tools where performance and memory safety decide the design: where I have used it since 2024 and with what.
The work I do with TypeScript: Model Context Protocol servers, Node.js services and where the type system earns its place in server-side code.
The work I do with RAG: parsing documents into embeddings, combining vector and full-text search, and the questions retrieval leaves unanswered.
The work I do with MCP: writing servers that open tools and context to a model, giving each project its own endpoint and treating every server as a trust boundary.
How I work with vector databases: choosing between pgvector, Milvus and Chroma, vector search inside PostgreSQL, and where they sit in a retrieval pipeline.
How I work with model serving: putting models behind an API with NVIDIA Dynamo-Triton, Ollama and FastAPI, and scaling them on Kubernetes.
I work with Apache Kafka: the message backbone of event-driven workflows and data pipelines, partition keys, consumer scaling and delivery guarantees.
I work with PostgreSQL: queue-table and index behaviour measurements, partial indexes, vector search with pgvector and relational store decisions.