The Workstation Blog
All Posts
Hands-on guides to AI infrastructure, MLOps, models, and shipping to production.
Large Language Models Explained: How LLMs Work and How to Run Your Own on Kubernetes
What are Large Language Models and how do they work? A clear, non-technical explainer for managers and engineers — tokens, embeddings, transformers, training and inference — plus production Kubernetes YAML to deploy your own LLM with Ollama and vLLM.
Balinder WaliaRead 
Can an AI System Process and File a Self Assessment Tax Return to HMRC Automatically?
Can AI file UK Self Assessment to HMRC? FileMyTax shows bank PDF → extract → classify → calculate → MTD with a human gate. Built by Workstation for CognoAI. Not tax advice.
Balinder WaliaRead 
Polyglot Benchmarks: Choosing the Right Tool for the Right Job
Stop default-stack bias. Polyglot Benchmarks compares njs, Lua, Python, Go, Rust, Bun, Java, and Kotlin on the same workloads — reproducible harness + live dashboard for architecture decisions.
Balinder WaliaRead 
Building a VAT Return System with Claude Code from Your Invoices and Expenses Database
I rebuilt my quarterly UK VAT return as a Claude Code project: read invoices + expenses from Postgres, classify each line by VAT treatment, compute HMRC Box 1 to 9 with a deterministic Python engine, dry-run, then submit to MTD with a human approval gate.
Balinder WaliaRead 
Unboxing Mac Studio M4 and Running Your First LLM
I unboxed a Mac Studio M4 Max with 128 GB unified memory and 40 GPU cores, installed Ollama, ran Llama 3.1 8B locally, and built a private RAG pipeline - all on one quiet, 6.48 W idle desk-side box.
Balinder WaliaRead 
NebulaDB vs Agent Memory Platforms: Why Open-Source Wins
A comprehensive comparison of NebulaDB — an open-source AI-native hybrid database written in Rust — against the category of agent memory platforms. Discover why NebulaDB's multi-protocol architecture, SQL-native vector functions, and streaming RAG pipeline make it the superior choice for most AI workloads.
Balinder WaliaRead 
Couchbase upgrades with CAO pause gates: a support-friendly overview
A support-friendly, SRE-ready overview of Couchbase Server rolling upgrades under the Couchbase Autonomous Operator (CAO), using spec.paused to gate swap-rebalance one node at a time—so you can stabilize, validate health, and keep rollback options open.
Balinder WaliaRead 
NebulaDNS Complete Feature Guide: k3s, Route53, Observability, and Enterprise DNS
Complete tour of NebulaDNS features: API-first control plane, Prometheus metrics, propagation gates, peer fingerprinting, DNSSEC roadmap, Helm/k3s operator story, and how to integrate with AWS Route 53 and CoreDNS.
Balinder WaliaRead NebulaDNS: Why Authoritative DNS Is a Game Changer in the AI Era
NebulaDNS is a safe-Rust authoritative DNS server with an API-first control plane, Prometheus metrics, propagation gates, and Kubernetes-native operations—built for visibility when AI services and agents depend on DNS more than ever.
Balinder WaliaRead 