Writing on AI Agent Systems | Carlos Paredes, NosisTech

I Ran a Good Model With Decent Hardware

May 28, 2026April 18, 2026 by Carlos Paredes

Bar chart comparing performance scores of local AI models including Llama 3, Mistral, Phi-3, and Gemma running on a mid-range consumer GPU.

I saw a benchmark showing Qwen3.6 outperforming Claude Sonnet on my laptop. I didn’t believe it. So I ran it myself — mistakes, PowerShell errors, and all.

I Pushed My Local AI’s Memory to the Limit

May 28, 2026April 18, 2026 by Carlos Paredes

Memory utilization dashboard showing VRAM, system RAM, swap space, and context window usage bars approaching capacity during a local AI inference stress test.

Testing context window limits on local AI — what worked, what slowed down, and what crashed.

Running Local AI on a Gaming Laptop: Set Up

May 28, 2026April 18, 2026 by Carlos Paredes

Screen of a gaming laptop showing terminal output of a local AI model loading via Ollama, with progress bars for model allocation and VRAM usage.

I set all of this up and do not fully understand everything I did. But it works, it is fast enough to be useful. That is the point of documenting it. The Hardware My machine is an Acer Predator PHN16S-71. CPU: Intel Core Ultra 9 275HX. GPU: NVIDIA RTX 5070 Ti with 12GB VRAM. RAM: … Read more

I Built My Own AI Server on a VPS

May 28, 2026April 18, 2026 by Carlos Paredes

Illustration of a glowing server rack hosted on a virtual private server, representing a self-hosted AI inference server with multiple GPU and storage layers.

I set up my own AI server on a Hostinger VPS running Ollama, proxying Claude through LiteLLM, and eventually adding Gemini. Here’s what that actually looked like.

How Transformer LLMs Work – What I Retained

May 28, 2026April 18, 2026 by Carlos Paredes

Architecture diagram of a transformer large language model showing the input, multi-head attention, feed-forward network, normalization, and output layers in sequence.

Transformer LLMs, reflection, and tool use — what I took away from a technical course as a non-developer.

I Made My Local AI Accessible From Anywhere

May 28, 2026April 18, 2026 by Carlos Paredes

Network diagram showing a laptop connected through an SSH or Cloudflare tunnel to provide remote access to a locally hosted AI model from any location.

I didn’t want to pay for a VPS just to access my local AI setup remotely. Cloudflare Tunnel solved it for free — and it took about ten minutes to set up.

VRAM vs RAM in Local AI Models

May 28, 2026April 18, 2026 by Carlos Paredes

Comparison diagram showing VRAM versus system RAM usage bars for local AI model inference, highlighting why GPU memory is the primary bottleneck.

When my AI model didn’t fit in GPU memory, it spilled into RAM. Here’s what that actually means for performance — explained without the jargon.

I Ran AI Models on My Laptop

May 28, 2026April 18, 2026 by Carlos Paredes

Bar chart visualization on a laptop screen showing inference performance of local AI models including Llama 3, Mistral, Phi-3, and Gemma running on consumer hardware.

I installed Ollama and two AI models on my Acer Predator laptop. It wasn’t smooth — but it worked. Here’s an honest look at what happened and what I learned along the way.

Agentic AI

May 28, 2026April 16, 2026 by Carlos Paredes

Completed the Agentic AI course by DeepLearning.AI. Key takeaways: multi-agent system design, reflection loops, tool use architecture, and what it all means for AI governance and compliance consulting.

AI for Everyone — For Business Leaders

May 28, 2026April 15, 2026 by Carlos Paredes

Diagram illustrating how AI capabilities connect to business leadership roles, showing a network of AI nodes linked to a central briefcase icon representing business strategy.

Completed AI for Everyone by Andrew Ng (DeepLearning.AI). Here’s what the course covers and why AI fundamentals matter for governance, compliance, and strategic decision-making.

One documented agent build in your inbox, every week