Johnson Lai · Forward Deployed Engineer 🇲🇾🇸🇬

🤖 Local LLM Enthusiast

Wed, 27 May 2026 01:32:00 +0800

DGX Spark RTX Pro 6000 RTX 3090

I run and optimize local LLMs for private, low-latency, cost-controlled AI workflows. My focus is practical deployment: inference speed, eval quality, tool calling, and fitting models into real teams.

What I work on Link to heading

Running models with llama.cpp, vLLM, and SGLang
Fine-tuning open models such as Qwen, Gemma, Llama, and Pi.dev variants
Building evals with lm-evaluation-harness, evalplus, and custom benchmarks
Optimizing quantization, batching, KV-cache, and throughput for real workloads
Connecting local models to agent tools, RAG pipelines, and coding workflows

Selected work Link to heading

Shipped DFlash support for Blackwell consumer GPUs in Luce.
Benchmarked Bonsai models on DGX Spark / GB10.
Run local junior-dev style agent workflows with Qwen and coding models.

If your team needs private local inference or local LLM workflow design, start with my Forward Deployed Engineer page.

⚡ Crypto × AI Engineer

Wed, 27 May 2026 01:31:00 +0800

I build at the intersection of Crypto AI and Crypto × AI: decentralized inference, on-chain AI agents, verifiable data, and agentic systems that need economic coordination.

Focus areas Link to heading

Decentralized inference: running and optimizing AI workloads across permissionless networks.
On-chain AI agents: identity, coordination, and transparent state for autonomous agents.
Agent commerce: job data, settlement flows, and reputation systems for AI services.
Verifiable AI: making model outputs, data, and decisions easier to inspect and challenge.

Relevant work Link to heading

Head of Data & AI at Chromia
Cofounder/CTO experience at Chasm Network
Writing on Chasm, Chromia, Virtuals ACP, Tee Hee, and AI agent transparency
Dune dashboards for on-chain analytics and protocol visibility

Selected writing Link to heading

🛠️ Forward Deployed Engineer Malaysia & Singapore

Wed, 27 May 2026 01:30:00 +0800

I’m Johnson Lai, a Forward Deployed Engineer in Malaysia/Singapore.

I help teams turn AI ideas into working systems: workflow automation, AI agents, local LLMs, RAG, evals, and production deployment. I work close to the team, understand the actual operations, and ship practical systems instead of slide decks. As a Forward Deployed Engineer Singapore and Forward Deployed Engineer Malaysia, I bridge teams across both locations to deliver production-ready AI systems.

I help with Link to heading

AI workflow automation
Internal AI agents and tool-use systems
Local/private LLM deployment
RAG, evals, and quality gates
AI-native product and operations builds

Track record Link to heading

Consultation and advisory for fintech and listed-company teams on practical AI deployment.
Head of Data & AI at Chromia, building agent-first workflows for marketing, engineering, and operations.
Local LLM and GPU workflow experience across llama.cpp, vLLM, SGLang, Qwen, evals, and coding agents.
Crypto × AI experience across decentralized inference, on-chain agents, and verifiable AI systems.

If you need a Forward Deployed Engineer, AI Solutions Architect, or hands-on AI consultant in Malaysia/Singapore, reach me on LinkedIn or Twitter/X.

My Multica workflow: a human + local-LLM agent fleet

Fri, 22 May 2026 20:00:00 +0800

Why agent-first Link to heading

Most teams still treat AI agents like fancy autocomplete: open IDE, ask Copilot, paste, repeat. That’s a tool, not a workflow. We saw the hype around tools like OpenClaw and Hermes Agent, but many users install them, poke around, and then get lost because they do not have an operating loop for using agents effectively.

As a Forward Deployed Engineer, my job is to redesign the workflow itself: agents handle the repeatable execution, humans review the highest-leverage decisions, and the system keeps shipping.

How to set up ClawChain

Wed, 11 Feb 2026 14:00:00 +0800

ClawChain is an on-chain social network on Chromia where only AI agents can post. Think of it as a social feed, but every account and every post lives on the blockchain.

Why does that matter? When Moltbook launched as an AI-only social network, a security researcher registered 1 million fake agents on it. There was no way to tell real agents from fakes, and the platform fell apart. ClawChain avoids this by requiring a wallet transaction to register and a verified X account to claim your agent, so spinning up fake accounts at scale isn’t really feasible.

Guide on Storing Virtual ACP data on Chromia

Thu, 20 Nov 2025 17:59:18 +0800

Virtuals’ ACP job data is recorded on Base chain, but on-chain data is hard to query, this is because it’s buried in calldata that’s tough to parse.

Most ACP agents today use centralized db for readability, but this approach sacrifices transparency and brings single points of failure.

What if you could have easy querying without losing decentralization or verifiability?

With just a few lines of code, you are able to use Chromia as a decentralized database for ACP agents, unlocking easy, dev-friendly queries while preserving transparency and simplicity.

XiaoZhi: The Open Source $15 AI Robot OS That's Changing Physical AI

Fri, 06 Jun 2025 09:59:18 +0800

Remember when the Rabbit R1 ($199) and Humane AI Pin ($700+subscription) promised to revolutionize AI hardware? They raised hundreds of millions, generated massive hype, and ultimately disappointed users with broken promises and locked ecosystems.

Rabbit R1 and Humane AI Pin

What if I told you there’s a $15 device that actually delivers what those expensive gadgets promised?

I discovered it during a recent trip to Shanghai, a viral AI physical robot called XiaoZhi (小智) that’s taking China by storm while remaining virtually unknown in the West. According to reports, XiaoZhi has already crossed the 100,000 active device threshold in just a few months, potentially making it the first AI-native hardware ecosystem to reach such numbers this quickly.

Chasm Story: Building Chasm Network Through the Bear Market

Mon, 04 Nov 2024 10:21:52 +0800

I left CoinGecko to start my own ventures into founding Hooga Gaming, which later pivoted into Chasm Network. After 3.5 years with Hooga Gaming & Chasm Network, I’m transitioning to an advisory role with Chasm Network while joining Chromia as Head of Data & AI. I’ll continue supporting the Chasm team as they build upon our established foundation.

Today is my last day with @CoinGecko, moving forward I will be building @HoogaGaming full time. Don’t think of this as a farewell, more towards that I’ve graduated from CoinGecko 🤗

Here’s a 🧵 about what I’ve built in this 3 years+ journey at CoinGecko 💚 pic.twitter.com/0WU3IMGKUu

Technical Analysis of Tee Hee: A TEE-Based Autonomous AI Agent

Wed, 30 Oct 2024 10:21:52 +0800

Introduction Link to heading

Tee Hee represents an intriguing implementation of a Trusted Execution Environment (TEE)-based AI agent operating on Twitter, which was just released today by Nous Research.

This is the first provably secure AI agent that prevents anyone from accessing the Twitter account after 7 days by using a TEE.

The article Setting Your Pet Rock Free by Nous Research covers how it works at a high level.

Let's talk about transparency in AI Agent

Sat, 26 Oct 2024 22:59:18 +0800

Recently, we’ve witnessed a wild rally of AI agent coins and AI memecoins across Crypto Twitter. But this raises an important question:

pic.twitter.com/B1cG6sfTmS
— Bold (@boldleonidas) October 22, 2024

We’re living in a bizarre timeline—one where we’re scrambling to prove something is actually AI rather than human. Think about it: we’ve spent decades dealing with CAPTCHAs and Turing tests to prove we’re human, and now we’re doing the complete opposite. 😂

As more AI agents pop up daily, we’re suddenly asking, “Wait, is this really an AI or just someone pretending to be one?” It’s both ironic and fascinating how the tables have turned, especially as these AI agents become more woven into our daily digital lives.

Chasm x Chromia: Paving the Way for Decentralized AI

Mon, 21 Oct 2024 10:21:52 +0800

Imagine a world where AI influences political elections, determine your online content, and decide your loan applications. As AI agents proliferate across platforms, a crucial question emerges: how can we verify whether we’re interacting with genuine AI systems? This isn’t science fiction – it’s our current reality, and blockchain technology, with its ability to timestamp and verify AI inferences on-chain, might help us solve this challenge.

Trust me Bro!

My Journey of Automating a Telegram Game

Wed, 09 Oct 2024 05:59:18 +0800

Automating boring tasks has always been a driving force behind my passion for building software.

A few months ago, I stumbled upon the Telegram game Lumberjack, which seemed like an ideal project for testing automation using simple pixel matching mechanics. Its straightforward mechanics made it an appealing target for experimenting with different automation approaches.

The Game: Lumberjack Link to heading

Lumberjack is a simple Telegram game where players chop a tree while avoiding branches. The objective is to press the arrow key opposite to the side of the last branch to avoid hitting the lumberjack’s head.

Blockchain IoT Weather Oracle on Chromia with Arduino

Sun, 22 Sep 2024 11:21:52 +0800

Introduction Link to heading

Integrating IoT with blockchain technology enhances data security, transparency, and reliability—crucial for IoT applications. Blockchain’s tamper-proof ledger and ability to timestamp data make it ideal for securely managing IoT information.

In this guide, we’ll build a blockchain-based IoT weather oracle using a thermistor sensor, Arduino, and Johnny-Five.js, storing real-time data on Chromia, a Layer-1 relational blockchain platform.

Project Overview Link to heading

We’ll create a weather data oracle using:

Run Chasm Node with Home PC

Sat, 20 Jul 2024 15:59:18 +0800

Chasm nodes are lightweight and can run on devices with as little as 2GB of RAM, including a Raspberry Pi. This guide is intended for those who want to try running a scout node for fun and to understand the Chasm Network better.

Important Note: For optimal security, low latency, and better performance, we strongly recommend using a Virtual Private Server (VPS) instead of a home PC to run your Chasm node. This guide is primarily for educational purposes and for those who want to experiment with running a node at home.

Chasm Scout Architecture

Fri, 19 Jul 2024 07:23:36 +0800

At Chasm, we believe the future of deAI is open-source.

Intern was told that the code for Scout is now open source.https://t.co/k2UvJXxTWO pic.twitter.com/ikImINE8So
— Chasm (@ChasmNetwork) July 19, 2024

We’re excited to announce that we have open-sourced the Chasm Scout codebase, which is aimed at democratizing access to Decentralized AI networks.

If you are new to operating nodes, no need to fret! We made sure Chasm Scout to be a very easy node to set up, so easy that even a 5-year old can do it. Here is a step-by-step guide on How to run Chasm Season 0 Scout Node. We promise it’ll only take 15-30 minutes of your time.

Chasm Network Scout Season 0 Hosting Tutorial

Thu, 18 Jul 2024 11:21:52 +0800

In this guide, I will walk you through the process of running Chasm Scout in Season 0 using Digital Ocean. Please note that you should always refer to the Official Guide from Chasm Network for the most up-to-date information.

I’ve chosen Digital Ocean for two main reasons:

It’s easy to use, especially for those new to cloud services.
Unlike AWS or GCP, it doesn’t require going through a complex DevOps process.

While Digital Ocean may be more expensive than some alternatives, it offers a user-friendly experience that’s ideal for this guide. For those seeking more cost-effective options, platforms like Linode, Contabo, Google GCP, AWS or Microsoft Azure can also be considered.

About Johnson Lai

Thu, 18 Jul 2024 08:20:18 +0800

Head of Data & AI · Qwen Model Ambassador Link to heading

I’m a Forward Deployed Engineer based in Malaysia/Singapore. I embed with teams to fill the gap between what AI tools can do and what the team actually needs, then ship production systems that run in their real environment instead of staying as slide decks.

Self-taught engineer and UWA alum. Today I’m Head of Data & AI at Chromia, where I create an agent-first approach to marketing, engineering, and operations. Also Model Ambassador at Qwen. Previously Cofounder/CTO at Chasm, COO/CTO at Hooga Gaming, and Senior R&D / Mobile Engineer at CoinGecko.

Johnson Lai · Forward Deployed Engineer 🇲🇾🇸🇬

🤖 Local LLM Enthusiast

What I work on Link to heading

Selected work Link to heading

⚡ Crypto × AI Engineer

Focus areas Link to heading

Relevant work Link to heading

Selected writing Link to heading

🛠️ Forward Deployed Engineer Malaysia & Singapore

I help with Link to heading

Track record Link to heading

Related pages Link to heading

My Multica workflow: a human + local-LLM agent fleet

Why agent-first Link to heading

How to set up ClawChain

Guide on Storing Virtual ACP data on Chromia

XiaoZhi: The Open Source $15 AI Robot OS That's Changing Physical AI

Chasm Story: Building Chasm Network Through the Bear Market

Technical Analysis of Tee Hee: A TEE-Based Autonomous AI Agent

Introduction Link to heading

Let's talk about transparency in AI Agent

Chasm x Chromia: Paving the Way for Decentralized AI

My Journey of Automating a Telegram Game

The Game: Lumberjack Link to heading

Blockchain IoT Weather Oracle on Chromia with Arduino

Introduction Link to heading

Project Overview Link to heading

Run Chasm Node with Home PC

Chasm Scout Architecture

Chasm Network Scout Season 0 Hosting Tutorial

About Johnson Lai

Head of Data & AI · Qwen Model Ambassador Link to heading