Thoughts on GenAI in production
Technical deep-dives, case studies, and field notes from building enterprise AI systems across Southeast Asia.
Showing 7 posts
All posts
Rebuilding the Organization for the Post-AI World
The traditional tech company playbook is broken. Here's how AI forced ArcFusion to rethink hiring, project management, sales, and org structure from the ground up.
Your AI Agent Is Failing Silently. Here's How to See It.
Standard monitoring sees a 200 and moves on. Here's how to instrument the tool call layer, wire Datadog alerts to Slack, and simulate failures end-to-end so your AI agent's silent degradation never reaches your clients.
The Agent That Learns: How We Built a Self-Improving AI Loop That Went from 86% to 99.8% Accuracy
Most AI agents plateau. We built an orchestrator that rewrites its own prompts when it gets things wrong — taking our document parsing accuracy from 86% to 99.8% without a single manual tweak after the loop started running.
How Can We Defend Against Prompt Hacking?
Prompt hacking is the #1 LLM security risk per OWASP 2025. Here's how the attacks work, and how to build a layered defense against them.
Using SIFT Feature Matching to QA Generative Product Placement
Generative models are great at vibes, but they have no concept of product integrity. Here's how we built a zero-shot QA layer using classical feature matching to catch placement failures before they reach a human reviewer.
The On-Premise LLM Reality Check: When Self-Hosting Makes Sense (and When It Doesn't)
Before committing to a seven-figure GPU infrastructure build, here's what on-premise LLM hosting actually costs — and when it's worth it.
Ready to move beyond pilots?
Tell us what you are trying to achieve.
We scope it, build it, and run it.