Posts tagged "LLM" | Alexandre Agius

AI ENGINEERING

27 October 2026 · 11 MIN READ

30 Failed Attempts and $21.87: The Output Cap of My Agent Factory Was Counting the Model's Thinking

At its identity step my agent build factory asks the model for three homepage mockups per site. For one site, five launches in a day produced 30 failed attempts, ten minutes of waiting each and $21.87 of model spend for nothing, while another site got its mockups the same afternoon. The cause was one number set months ago: a 32,000-token cap on each answer, on a model that now reasons before it writes, and whose reasoning counts against that cap. Here is how I found it, the per-model cap that fixed it, the replay that proved it before the merge, and the second defect the same morning exposed: an e-mail the page promised and nothing sent.

Read → ∴⁂ #Agentic AI

CLOUD

23 July 2026 · 5 MIN READ

AWS + Cerebras: wafer-scale inference is coming to Bedrock

AWS is deploying Cerebras CS-3 systems in its data centers, pairing Trainium for prefill with the Wafer Scale Engine 3 for decode. Why disaggregated inference is the right architecture, and what makes a 4-trillion-transistor chip the right tool for the decode problem.

Read → ∴ #AWS

23 July 2026 · 5 MIN READ

Kimi K3 — the first open 3T-class model, explained

Moonshot AI shipped a 2.8-trillion-parameter open-weight model with a 1M-token context, native vision, and a new attention stack. What's actually new in the architecture, what the benchmarks say, and why it matters for builders.

Read → ∴ #AI

10 March 2026 · 14 MIN READ

Building a RAG System That Actually Works: Chunking, Vector Engines, and Testing

Most RAG tutorials stop at 'put vectors in a database.' This post covers what actually determines quality: how you chunk documents, which vector search engine to pick, and how to measure and iterate on retrieval performance using Bedrock Knowledge Bases and LLM-as-judge evaluation.

Read → ∴⁂ #AWS

01 March 2026 · 9 MIN READ

World Monitor: How Open-Source OSINT Is Democratizing Global Intelligence

A deep dive into World Monitor — an open-source intelligence dashboard that aggregates 150+ feeds, 40+ geospatial layers, and AI-powered analysis into a real-time situational awareness platform. What OSINT is, how these platforms work under the hood, and why it matters now more than ever.

Read → ∴† #OSINT

26 February 2026 · 17 MIN READ

LLM Architecture Explained Simply: 10 Questions From Prompt to Token

A beginner-friendly walkthrough of how an LLM actually works end-to-end: from typing a prompt to receiving a response — covering tokenization, embeddings, Transformer layers, KV cache, the training loop, embeddings for search, and why decoder-only models won.

Read → ∴⁂ #AI

26 February 2026 · 13 MIN READ

LLM Inference Demystified: PagedAttention, KV Cache, MoE & Continuous Batching

The 5 key concepts every cloud architect should know about LLM serving: PagedAttention, KV cache mechanics, continuous batching, MoE trade-offs, and real production numbers.

Read → ∴ #AI

25 February 2026 · 9 MIN READ

LLM Distillation vs Quantization: Making Models Smaller, Smarter, Cheaper

Two strategies to shrink LLMs — one compresses weights, the other transfers knowledge. A practical guide to distillation and quantization: when to use each, how to implement them with Hugging Face, and why the real answer is both.

Read → ∴ #AI

24 February 2026 · 9 MIN READ

Getting Hands-On with Mistral AI: From API to Self-Hosted in One Afternoon

A practical walkthrough of two paths to working with Mistral — the managed API for fast prototyping and self-hosted deployment for full control — with real code covering prompting, model selection, function calling, RAG, and INT8 quantization.

Read → ∴⁂ #AI

24 February 2026 · 14 MIN READ

Python, Transformers, and SageMaker: A Practical Guide for Cloud Engineers

Everything a cloud/AWS engineer needs to know about Python, the Hugging Face Transformers framework, SageMaker integration, quantization, CUDA, and AWS Inferentia — without being a data scientist.

Read → ∴❖ #AWS

24 February 2026 · 9 MIN READ

TFLOPS: The GPU Metric Every AI Engineer Should Understand

What TFLOPS actually measures, why FP16 matters for LLMs, and why the most important GPU bottleneck for inference isn't compute at all.

Read → ∴ #AI

23 February 2026 · 15 MIN READ

Transformer Anatomy: Attention + FFN Demystified

A deep dive into the Transformer architecture — how attention connects tokens and why the Feed-Forward Network is the real brain of the model. Plus the key to understanding Mixture of Experts (MoE).

Read → ∴ #AI

22 February 2026 · 11 MIN READ

Fine-Tuning Mistral with Transformers and Serving with vLLM on AWS

End-to-end guide: fine-tune Mistral models with LoRA using Hugging Face Transformers, then deploy at scale with vLLM on AWS — from training to production serving on SageMaker, ECS, or Bedrock.

Read → ∴❖ #AWS

09 February 2026 · 9 MIN READ

How LLMs Learn to Behave: RLHF, Reward Models, and the Alignment Problem

A practical walkthrough of how large language models are aligned with human values — from collecting feedback to PPO optimization and the reward hacking pitfalls.

Read → ∴ #AI

07 February 2026 · 8 MIN READ

A Practical Guide to Fine-Tuning LLMs: From Full Training to LoRA

Understand how LLM fine-tuning works, when to use it, and how to choose between full fine-tuning, LoRA, soft prompts, and other PEFT methods.

Read → ∴ #AI

All topics Full journal