Marcio Cunha

Technical Articles

Practical articles, in-depth tutorials, and architectural insights on software engineering, AI, and systems.

All categoriesArquiteturaAutomaçãoBackendCarreiraDevOpsEletrônicaEngenhariaFrontendHardwareIANegóciosProdutividadeRedes
IAHardware•4 min

Local Multimodal Model Inference Implementation with Dynamic VRAM Offloading

Learn how to run multimodal models locally by intelligently managing video memory and system RAM. Discover practical offloading techniques to prevent out-of-memory errors when processing images and text.

IAArquitetura•3 min

Implementation of Distributed Language Model Inference via Layer Splitting

Learn how to slice massive neural networks across different computers to run heavy artificial intelligence models without relying on centralized high-end hardware.

IAHardware•5 min

Optimizing Generative Models at the Edge with Dynamic Weight Quantization

Learn how to run heavy generative artificial intelligence on resource-constrained hardware using dynamic weight quantization techniques to reduce memory consumption without losing accuracy.

IAArquitetura•5 min

Implementing Autonomous Agents with Hybrid RAG in Distributed Vector Databases

Learn how to build efficient autonomous agents by combining hybrid data retrieval and distributed vector databases for large-scale semantic search.

IAArquitetura•4 min

Mitigating Language Model Hallucinations with Knowledge Graphs

Learn how to combine Large Language Models with Knowledge Graphs to perform real-time fact-checking and eliminate fabricated responses.

DevOpsIA•5 min

Deploy Pipeline Failure Mitigation via Proactive Error Log Analysis Using Neural Networks

Learn how to integrate neural networks to analyze error logs in real-time and prevent catastrophic failures in continuous delivery pipelines.

IAHardware•4 min

Local Language Model Quantization with Hybrid CPU and NPU Execution in Workstations

Learn how to run artificial intelligence locally using central processors and neural units combined with quantized models for maximum energy efficiency.

IAArquitetura•5 min

Distributed Language Model Inference with Tensor Parallelism in Heterogeneous Clusters

Learn how to run massive artificial intelligence models by splitting computational layers across different computers, overcoming hardware barriers and cutting costs.

IAHardware•4 min

Artificial Intelligence Model Acceleration on Edge Devices with Four Bit Quantization

Explore how compressing artificial intelligence models down to four bits enables complex neural networks to run directly on compact, offline hardware.

IAEngenharia•3 min

Efficient Fine-Tuning of Large Language Models with Low-Bit Quantization and LoRA

Learn how to adapt massive artificial intelligence models using less computational memory through LoRA adapters and low-bit quantization.