About · AI Architect · Malaysia

Who is Lindsey?

Lindsey is a full stack marketer turned AI architect based in Malaysia. She builds AI led systems that raise output while cutting cost, delivered as local scripts, company Apps with ERP or CRM, and multi agent teams. Every system runs on your own data, at controlled cost, built and managed by AI.

How Lindsey runs an entire marketing team on her own

I do not have a team. I built one.

Xiao An Orchestration and task routing, across the whole line

  1. 1Strategy and topics

    Brand strategy lead
    Brand strategy

  2. 2Content production

    Copy lead
    Ad copy
    Design lead
    Graphic design
    Video editor
    Video editing

  3. 3Distribution and reach

    Social lead
    Social media
    Visibility lead
    SEO and GEO

  4. 4Ads and growth

    Media buyer
    Ad operations
    Data lead
    Data analysis

Review lead · content review lead · build engineer · review engineer

Quality review · content review and compliance · system development · code review

Lindsey. An entire marketing team.

13 AI agents working together, turning a full day of repetitive operations into a short automated run.

AI Architect: Technical Capabilities

What sits under the team

Most people who can wire up Claude are engineers who don't know marketing; most who know marketing can't build the agents. I sit on both sides. I run the growth work and build the system that runs it.

1. Agent system architecture

Multi agent system design, handoff orchestration, and model driven dynamic routing. Capability boundary governance, task decomposition and dispatch, context compression, and the design of memory and lessons libraries.

2. AI engineering

Built on the Claude Agent SDK, defining agents and tools, and shipping my own MCP (Model Context Protocol) servers to expose local capabilities as standard tools. Prompt and context engineering, hallucination control and confidence calibration, structured tool I/O contracts with self healing retries, and human in the loop intervention points.

3. Agent guardrails & security architecture

A deterministic authorization kernel where every tool call, including every sub agent, passes one enforcement chokepoint, classified read / write / danger with a hard floor no autonomous mode can override. Policy lives in code, not in a prompt, so a manipulated model can't talk its way past it: data exfiltration prevention, command chain and URL smuggling defense, secrets isolation, filesystem sandboxing, and RCE / CSRF hardening.

4. Security assurance & red teaming

A formal threat model for the whole autonomous system: trust boundaries, a T1 to T6 threat matrix mapped to code, and honestly tracked residual risks. Backed by an adversarial red team test suite (injection, exfiltration, sandbox escape, secret reads) that asserts every attack is refused.

5. Retrieval & memory engineering

A hybrid search engine written from scratch: Okapi BM25 over an inverted index, dense embeddings, Reciprocal Rank Fusion, and a deterministic reranker, CJK aware, running on a production scale corpus. Layered, token budgeted context assembly, per agent memory scoping with PII redaction, and autonomous nightly memory consolidation gated by human review.

6. Evaluation & quality engineering

Measurement driven, not vibes driven. RAG evals (recall@k, MRR, faithfulness / groundedness, dense vs hybrid A/B), an LLM as a judge harness scoring agent behavior against rubrics and hard red lines, and a golden regression suite, all wired into CI as pass / fail gates.

7. Platforms & data pipelines

Day to day operation across TikTok Shop, TikTok Seller Center, Kalodata, Shopee, Meta Ads, and Google Business Profile. Cross platform pipelines that move that data into the systems I build, including browser session extraction that works against logged in sessions with zero LLM tokens and no dependency on official APIs.

8. Full stack delivery & deployment

End to end: from a framework free, zero dependency Node backend and native front end, to a native macOS app (Swift / WKWebView) with process self healing, to reproducible Docker builds and JAMstack edge deployment (Astro + Cloudflare Workers + GitHub Actions). Including source code protection: compiling to V8 bytecode as a dual artifact release for IP safety.

9. Media, GEO tooling & content systems

FFmpeg, Whisper, text to speech, and local image to video models. Built my own GEO / AEO auditing engine, citability scoring and AI crawler checks, plus multimodal ad authenticity and compliance checking, and Claude Skills authoring at scale (90+ shipped).

Architecture direction

Local first deployment, execution layer separated from the judgment layer, information barriers between company and personal data, and compliance minded designs where data stays inside the network.

Marketing & Growth

The operator behind the architect

I didn't come to AI from computer science. I came from running paid media, content, and e commerce by hand, which is why the systems I build actually fit how the work gets done.

1. Paid media & ads management

Performance marketing across TikTok Ads, Meta Ads, Google Ads, Shopee Ads, and Lazada Ads, plus TikTok Shop GMV Max. Campaign strategy, creative testing, budget pacing, and ROAS optimization, managing ad spend across a portfolio of e commerce clients at a consistently profitable blended ROI.

2. E commerce operations

Day to day operation of TikTok Shop, Shopee, and Lazada stores, listings, GMV growth, KOL and affiliate programs, and brand protection (removed a large batch of counterfeit SKUs). Scaled one Shopee store from a modest baseline to a substantially higher monthly revenue.

3. Content & copywriting

Trilingual content, English, 中文, Bahasa Malaysia. Hooks, angles, and platform native copy; deconstructing top performing posts into reusable patterns, then writing from them.

4. Video production

Shooting and editing short form video (CapCut / JianYing) for TikTok and social, script to storyboard to final cut, plus AI assisted generation where it speeds the work.

5. SEO / GEO / AEO

Search and AI search optimization, content strategy, Google Business Profile, Semrush, and E-E-A-T signals. Built the company's first GEO / SEO capability from zero and ran it end to end.

6. Social & web presence

Social media management across platforms, plus website design, owning the full funnel from discovery to the page that converts.

I spent years in hands on in house marketing and e commerce before moving into AI architecture. Today I design AI systems across four layers, planning, tools, memory, and state, so a small team can run like a large one. The work is not a demo. It runs in production, every day, on real company data.