Articles › AI

Your team downloaded Llama 4, spun up a GPU box, and got a chatbot answering questions in an afternoon. That part is genuinely easy in 2026 — open-weight models are good, quantization tooling is mature, and inference frameworks like vLLM are well-documented.

This isn’t a hypothetical. It’s the single most common architectural mistake in enterprise RAG deployments, and it’s almost never caught in a demo. Demos use one document set and one user. Production has HR records, legal contracts, board decks, and 200 employees — and the moment two of those employees have different permissions on the […]

From "why go local" to "what do I actually buy"

A 25-point self-assessment for IT and security leaders in Southeast Asia and Japan

title: "Why Enterprises in Southeast Asia and Japan Your teams are already using AI. The question is whether you know about it.

In the previous post in this series, we built a streaming AI chatbot backed by a FastAPI server. That covers the majority of enterprise use cases. But for a growing set of requirements — strict data residency, offline functionality, zero per-token cost at scale — the right answer is to run the model directly on […]

Most React Native tutorials stop at the UI layer. They show you how to render chat bubbles and handle keyboard offsets—then hand-wave the backend with a vague "call the OpenAI API from your app."

If you’re building a RAG pipeline or any application that needs semantic search, you’ll eventually need to decide where to store your embeddings. Dedicated vector databases (Pinecone, Qdrant, Weaviate) are one option. But for most teams — especially those already running PostgreSQL — pgvector is the faster, cheaper, and operationally simpler path.

Most organisations that evaluate AI for internal use eventually ask the same question: do we use ChatGPT, or do we deploy something ourselves?

Every enterprise that’s tried ChatGPT for internal knowledge work hits the same wall: the model doesn’t know your products, your policies, your contracts, or your procedures. It hallucinates. It answers confidently with the wrong information.

There’s a graveyard somewhere in every enterprise IT department. It’s full of demos that wowed a steering committee in 2024, RAG prototypes that beat ChatGPT on three cherry-picked questions, and copilots that looked unstoppable in a sandbox. Then someone asked the hard question — can we put this in front of 4,000 employees, integrated with […]

If you’re running ERPNext across multiple Asian factory operations — say a manufacturing plant in India, a Southeast Asian satellite (Thailand, Vietnam, Indonesia), and an export-oriented Chinese subsidiary — you’ve probably noticed the same pattern in three different forms. ERPNext handles the manufacturing operations beautifully. The Manufacturing module, the BOM/work order/subcontracting flow, the multi-warehouse stock […]

If you run Odoo across multiple Asian markets — say, a manufacturing operation in Thailand, a sales subsidiary in Japan, and a procurement office in China — you’ve probably discovered the same thing in three different ways. The standard invoice digitization feature works beautifully on the AWS bills and the European supplier invoices. It struggles, […]

Enterprise ERP integration is a solved problem — at least, that’s what your SI vendor will tell you. The reality inside tier-2 and tier-3 manufacturing plants is very different: brittle point-to-point connectors, hand-rolled ABAP scripts, and decades-old middleware that breaks every time a vendor pushes a schema update. The data exists. The systems exist. The […]

Every vendor with a security product has bolted "AI" onto the marketing page in the last eighteen months. Most of it is rebranded ML classification: anomaly detection that already existed, dressed in 2024 clothes. Where it gets genuinely interesting — and where most teams fail — is when you actually wire a tool-using LLM agent […]

Walk into any mid-sized accounting firm in Bangkok and you’ll see the same thing. Three monitors. One has FlowAccount open — logged in as the client. Another has Excel — the master tracker, color-coded by partner, deadline, and filing type. The third has LINE — twelve group chats, one per client, full of photographed receipts, […]

How much RAM, VRAM, and GPU do you actually need? An engineer’s guide to picking hardware for running LLMs locally — without overspending, and without surprises.

title: "Why Your Finance Team Spends 40% of Their Week on Work AI Can Now Do" For CFOs and COOs of mid-market Thai manufacturers and distributors If you run finance or operations at a 100 to 500 person manufacturer or distributor in Thailand, you already know the shape of the problem. Your team is technically […]

When your AI system goes live, it doesn’t just gain capabilities — it gains an attack surface that didn’t exist before. Most enterprises have invested years hardening their applications, networks, and endpoints. But the AI layer introduces a fundamentally different category of vulnerability that traditional security tools were never designed to handle.

You’ve already tuned top_p, top_k, and repeat_penalty. Your output stopped looping and the nonsense dropped. But your coding model still wanders off-topic, forgets earlier code, or refuses to stop where you want it to.