Insights | Faizan Ahmed Raza | Faizan Ahmed Raza

Insights

Thinking Out Loud

Practical articles on AI engineering, agentic systems, cloud infrastructure, security, and building software that actually ships.

Topics

Showing 20 of 20 articles

AI SecurityOct 5, 2026

AI Is Flooding Security Teams and Gaining Desktop Access: What October 2026's Headlines Mean for Builders

Google paused its open-source bug bounty after a flood of automated reports, GitLab patched a critical sandbox escape in its AI Gateway, and Gemini is reportedly testing full desktop control. Three stories, one pattern: AI is changing both who finds vulnerabilities and what an exploit can reach.

6 min readRead

AI SecurityAI AgentsBug Bounty

Responsible AISep 26, 2026

Responsible AI in Practice: What the EU AI Act, NIST, and Frontier Labs Actually Require in 2026

"Responsible AI" spent years as a slide in a slide deck. In 2026 it's become a compliance deadline, a procurement requirement, and — for at least one frontier lab — a versioned policy document with its own changelog. Here's what's actually binding, what's voluntary, and what to do about it.

6 min readRead

Responsible AIAI GovernanceEU AI Act

AI SecuritySep 26, 2026

AI Supply Chain Security: Malicious Models and Hallucinated Packages Are the New Attack Surface

Software supply chain security became a boardroom topic after SolarWinds. The AI stack has its own equivalent — and it's already being probed in the wild, through poisoned model weights on Hugging Face and package names that AI coding assistants simply make up.

6 min readRead

AI SecuritySupply ChainMachine Learning

AI SecuritySep 25, 2026

Prompt Injection: Why Your AI Agent Is Only as Secure as Its Weakest Tool Call

Prompt injection is quickly becoming the SQL injection of the LLM era — except instead of corrupting a database query, it can hijack an agent that has real tools. Here's what it actually is, why agents make it worse, and how to defend against it.

6 min readRead

AI SecurityPrompt InjectionLLMs

Cloud ComputingSep 22, 2026

FinOps for AI: Getting Control of Your Cloud Bill Before It Controls You

AI workloads turn cloud bills into a moving target — GPU instances, autoscaling inference, and forgotten dev environments add up fast. Here's a practical FinOps framework for keeping AI infrastructure spend under control without slowing teams down.

5 min readRead

FinOpsCloud ComputingAI Infrastructure

AI/ML EngineeringSep 14, 2026

The Real Cost of Training Deep Learning Models at Scale

Everyone quotes the headline number for training a frontier model, but almost nobody explains where the money actually goes. Here's a practical breakdown of compute, data, and the hidden costs that decide whether a training run is worth it.

5 min readRead

Deep LearningGPUsMachine Learning

AI AutomationSep 5, 2026

Multi-Agent AI Systems: When One Agent Isn't Enough

Single-agent demos look great until the task gets complex. Here's why teams are moving to multi-agent architectures, the coordination problems nobody shows in the demo, and how to avoid building a system that argues with itself.

5 min readRead

AI AgentsMulti-Agent SystemsOrchestration

Robotics & AIAug 28, 2026

Physical AI: How Foundation Models Are Finally Making Robots Useful

For decades robots needed a human to hand-code every motion. Foundation models trained on video, simulation, and teleoperation data are changing that. Here's what's real, what's still a demo reel, and why it matters beyond the hype.

5 min readRead

RoboticsPhysical AIFoundation Models

AI/ML EngineeringAug 15, 2026

Tokens & Credits Explained: Why AI Is Priced the Way It Is

Every AI product bills you in "tokens" or "credits" — and it's rarely obvious what you're actually paying for. Here's how tokenization works, why it drives cost, and how to use it to your advantage.

5 min readRead

LLMsTokensAI Costs

Editorial & Research Transparency

Articles are authored by Faizan Ahmed Raza, backed by production engineering experience, reproducible code, and peer-reviewed citations.

Editorial Standards →