The Loop Asia

"Everybody should be working to replace themselves right now." Buu Lam's blunt career advice — and the reasoning behind why F5 built a completely different architecture for AI gateways than API gateways.

Buu Lam is Director of Community Evangelism at F5, running the 20-year-old F5 DevCentral community and now focused on the shift from traditional API gateways to AI gateways.

What we cover:

• Why an AI gateway needed a different architecture — WebAssembly processors instead of event-based scripting

• Why AI gateways have to inspect both request AND response, not just signatures

• A live demo of a prompt injection game where simple adjectives bypass system prompts

• The agency question: how much should a security function autonomously halt a transaction flow?

• Why staying valuable means actively trying to automate your own job first

The Loop Asia: www.theloop.asia

"You only have a platform if people build something on top of it you didn't count on." Erik Wilde on why standards aren't the opposite of innovation — they're the precondition for it.

Erik Wilde is an ambassador for the OpenAPI Initiative, spreading the standard's value across the global apidays conference circuit.

What we cover:

• The double-sided pyramid: infrastructure at the bottom, a standardized platform in the middle, innovation on top

• The 100-year automotive platform analogy — entire model lineups from a handful of shared platforms

• Why Uber Eats likely took a fraction of the effort by reusing Uber's existing services

• Why there's no single "true" customer model across a large organization

• Where the OpenAPI conversation is heading: scaling API practice, not explaining what an API is

The Loop Asia: www.theloop.asia

Singapore built the perfect launchpad for startups — and that's part of the problem. Jed Ng on why comfort is often the enemy of the market-expansion decisions founders need to make early.

Jed Ng helped build RapidAPI into the world's largest API marketplace before turning angel investor. He now runs a syndicate of 1,300+ investors and Angel School, an accelerator for angel investors and emerging syndicate leads.

What we cover:

• Why every investment thesis needs a clear exit plan before market sizing

• The "forcing function" of small home markets, and why Singapore's advantages can become a trap

• Local control risk: why founders need to internalize market entry rather than depend on local partners

• How to test a new market cheaply with partners before doubling down

• What startups eyeing Asia get wrong about "the natural corridor"

The Loop Asia: www.theloop.asia

"Language models isn't AI" — Uli Hitzel's blunt correction to how most teams talk about GenAI, and the starting point for actually using it well in software development.

Uli Hitzel is an AI practitioner and instructor at NUS, with a hands-on career spanning Microsoft, Axway, and Dyson before focusing full-time on applying generative AI tools in the enterprise.

What we cover:

• Why you don't need to understand how a language model works to get real value from it

• The junior-developer trap: producing code from AI you can't debug or explain

• Where GenAI genuinely helps across the SDLC — architecture, documentation, tests, diagramming

• Why enterprise teams need to plan for tool and vendor churn

• The "cloud adoption" analogy for AI rollout: start non-critical, build trust

The Loop Asia: www.theloop.asia

A dashboard vendor can't discount their way out of a 30-minute mitigation window they just broke. Buu Lam on why observability tooling decisions are C-level problems disguised as SOC-level ones.

Buu Lam is Community Evangelist at F5, working with organizations navigating multi-cloud infrastructure, DevOps resilience, and security operations.

What we cover:

• Why vendors that can't integrate into an existing ecosystem eventually stall or get acquired

• The hidden cost of a tool that doesn't plug into your existing dashboard

• Behavioral DDoS signals — how a slowdown can be a red herring for something else

• Why resilience decisions always come down to multiple data points, never one signal

• The tier-1 SOC analyst problem: pattern recognition experienced operators have and new ones don't

The Loop Asia: www.theloop.asia

A top-50 global bank let Chuck Herrin's team print money undetected — valid login, valid token, API responding exactly as designed. No hacking required, just an exposed API nobody was watching.

Chuck Herrin is Field CISO at F5, following 20 years as CISO at global financial services firms including AIG and Texas Capital Bank, and CTO of API security firm Wib (acquired by F5).

What we cover:

• Why 92% of the attacks F5 detects on its global network target API endpoints, not web apps

• The real bank breach case study — a "salami slicing" attack undetected across hundreds of millions of transactions

• Why API attacks are often a single request — a "kill shot," not a classic kill chain

• API discovery in a multi-cloud world: why no single lens shows every API you're exposing

• How AI agents introduce new risks — reward-seeking bots gaming their own metrics

The Loop Asia: www.theloop.asia

Kubernetes debugging is a deep iceberg — pods, services, ingress, persistent volumes, each layer its own body of knowledge. Nilesh Gule shows how generative AI can compress that whole diagnostic flowchart into plain English.

Nilesh Gule is an Enterprise Architect, Microsoft Most Valuable Professional, and Docker Captain, working with Kubernetes since 2017.

What we cover:

• K8sGPT — the open-source tool that scans your cluster and diagnoses issues in plain English (18+ languages)

• Why it's built to run as a Kubernetes operator, continuously, not a command you remember to run

• The backends it supports — OpenAI, Azure OpenAI, Ollama for local models, Bedrock, Gemini, and more

• Why Nilesh keeps it suggestion-based, not self-healing

• How it integrates with Trivy, Prometheus, Keda, and Kyverno

The Loop Asia: www.theloop.asia

"You build it, you run it" is the DevOps aspiration every org repeats. Nilesh Gule has spent 21 years watching what actually happens between that slogan and reality — and where it goes next.

Nilesh Gule is an Enterprise Architect, Microsoft Most Valuable Professional, and Docker Captain, with two decades of hands-on DevOps experience.

What we cover:

• Why DevOps implementation depends more on organizational structure than on tooling

• "Shift left" in practice: moving testing and security earlier in the development cycle

• Two real models for structuring DevOps teams — shared services vs. embedded SRE

• The evolution from CI, to DevOps, to SRE, to platform engineering

• Why starting a project without CI/CD baked in has become almost unthinkable

The Loop Asia: www.theloop.asia

Why does Koo Ping Shung insist on "trustable AI" instead of the industry-standard "trustworthy AI"? Because you can only call a human trustworthy — an object is, at best, trustable at a given task.

Koo Ping Shung is Practicum Director at Data Science Rex, and in Part 2 he moves from governance theory into what actually happens inside organizations trying to build it.

Part 2 of 2 — what we cover:

• Why most organizations haven't reached the maturity to attempt real AI governance yet

• Documentation as the actual foundation of governance — LLMs are only as good as what trains them

• Using LLMs to help new hires document and onboard faster

• Why Koo doesn't believe in "ethical machines"

• The real distinction between "trustable" (task-level) and "trustworthy" (entity-level) AI

Missed Part 1? It covers AI bias, governance committee structure, and AI literacy.

The Loop Asia: www.theloop.asia

Bias in AI can never be fully removed — because it comes from us. Koo Ping Shung explains why that means governance has to be about awareness and transparency, not the impossible goal of a "bias-free" model.

Koo Ping Shung is Practicum Director at Data Science Rex, with two decades in data science spanning the shift from business intelligence to analytics to AI.

Part 1 of 2 — what we cover:

• Why bias in AI can be reduced but never fully eliminated

• The five disciplines an AI governance committee actually needs

• Why "AI literacy" doesn't mean becoming a data scientist

• The car-dashboard analogy for trusting a system you can't fully understand

• Why companies need a real feedback channel for AI decisions

Part 2 continues with data practices, documentation, and API governance.

The Loop Asia: www.theloop.asia