Stuck on AI or automation? I can try to help.
Oscar Developments is one person who builds software and runs AI tools on home servers. I'm not an agency or a certified consultant. If your problem overlaps with what I know, I'll do my best to help.
Or email [email protected]
What I have hands-on experience with
- Running language models locally on my own GPU server
- Self-hosting on Proxmox and Linux, with backups
- Scripting in Python and PowerShell
- Using Claude, Gemini, and ChatGPT day to day
- Calling AI models through APIs, including OpenRouter
- Setting up MCP connectors so assistants can use tools
- Building desktop and Android apps
All of it comes from my own projects, not years of client work.
Where my experience comes from
Everything I know, I learned by building my own things. I write and ship open-source apps, and I run a small server setup at home that hosts local AI models, my websites, and nightly backups.
That is real, hands-on experience, and it is also the limit of it. I haven't worked inside big companies, and I won't pretend I have.
If you message me, you are talking to the person who would actually do the work.
The AI side
The tools I actually use, and what I can help with in each.
Claude, Gemini, and ChatGPT
I use all three regularly and know where each one is stronger. I can help you pick one for a task, set up projects and custom instructions, and write prompts that get useful answers.
AI APIs
I've built apps and scripts that call models through APIs, including OpenRouter. I can help with API keys, choosing models, keeping costs down, and handling errors and rate limits.
MCP connectors
MCP, the Model Context Protocol, lets assistants like Claude use outside tools and data. I've set up connectors that let an assistant run commands on my servers, with safeguards in place.
Local models
I run open models with Ollama on my own GPU server. That is the option when your data can't go to a cloud AI service.
Being upfront
What you can and can't expect from me.
No promises
I can't guarantee results. I'll tell you what I think will work, what might not, and when something is outside what I know.
No fixed pipeline
I don't sell packages. How I work depends on your team: the tasks you do, the tools you already use, and what you actually need.
Messages are free
Ask questions and go back and forth as long as you need. Nothing costs anything until we agree on a written plan.
Things I can try to help with
Starting points, not services. If your problem is close to one of these, it's worth a message.
Looking at a workflow
Walk me through how a task gets done. I'll share where a script or an AI model might save time, and where it probably won't.
Small automations
Scripts and simple pipelines that take repetitive steps off your plate, including ones that call Claude, Gemini, or ChatGPT through their APIs.
Local AI setups
Help choosing hardware and getting language models running on your own machines, based on what I run myself.
Explaining what we built
Plain notes and a walkthrough, so you understand your setup and aren't stuck depending on me.
If we work together
Roughly how it goes. Everything inside these steps depends on you.
Talk, free
Messages back and forth about the problem. No charge, no commitment.
Honest answer
If it looks like something I can help with, I write down what I'll try and what it costs. If not, I say so.
Try it
Work starts only after you agree to the plan. You see progress along the way.
Hand off
Notes and a walkthrough of whatever we set up.
Not sure if I can help?
Ask anyway at [email protected]. Messages are free, and I'll be honest if it's outside what I know.