AWS ML Hero. MLE @boltapp. Failed Chemist. Blogging about my ML/DL journey. IceVision core-dev.
Belgium
Joined November 2011
- Tweets750
- Following266
- Followers2K
- Likes1.7K
Pinned Tweet
If you have trained a model, you want to deploy it to SageMaker, and your first instinct is to start screaming, please😌 and keep reading.
Today, I am launching a course: How to deploy an ML model to SageMaker with a custom Docker container
👉 Get it here: frapochetti.gumroad.com/l/md…
I finally shipped Visual Neurons: a personal AI studio for image/video generation.
9 models. One interface. Zero lines of code written by me.
Claude Sonnet 4.5 + Cursor built it.
📖 Full story: francescopochetti.com/buildi…
🎥 Demo: youtu.be/1TS7quWMmIY
Here's what it took 🧵
3 complete pivots:
Feb: AWS Amplify (15-min builds, mysterious failures)
Apr: Vercel + Supabase (too many platforms, blind LLM debugging)
Aug: EC2 + local dev → shipped
The breakthrough: KISS. No S3, no complexity. Just Next.js + SQLite + nginx.
The full story covers: disasters, pivots, technical architecture, and what actually works when vibe-coding with LLMs.
Tech: Next.js 15 • React 19 • Prisma • Replicate • AWS Bedrock
#AI #LLMs #VibeCoding #BuildInPublic
Working on adding Runway Gen-4 video generation.
Close but not there yet.
runwayml.com/research/introd…
When I thought I was done with outpainting, I started stumbling on frontend timeouts. The StabilityAI API call is successful though (I see it in 💸). Lambda logs confirm too.
Seems AppSync plays tricks on me. Here's how ChatGPT puts it 👇 (and yes, I have to move to async)
```
If you are calling the Stability outpainting API from an AWS Lambda resolver (via AppSync), remember that:
AppSync imposes a 30-s timeout for all resolver calls. Even though your Lambda function can be configured for a 120-s timeout, AppSync itself will cut off at 30 seconds.
Managed to add Outpainting from StabilityAI 🎉
Now hardcoding 100px in each direction.
platform.stability.ai/docs/a…
Integrated StabilityAI image generation.
After the 20-commit-3-days refactor it was a breeze 🙌
It boiled down to implementing the `generateImage` func inside the `stabilityProvider.ts` and the dispatcher propagates to Lambda and the frontend.
github.com/FraPochetti/visua…
Not sure what they are doing at @AnthropicAI, but Claude is basically unusable for any reasonably long context.
ChatGPT handles the same queries without any issues.
How do you convert a codebase to text for LLM consumption?
1️⃣ replace github with gitingest gitingest.com/fastai/fastai
2️⃣ replace g with u uithub.com/fastai/fastai
3️⃣ clone the repo and write a script to only export specific chunks of interest.
Details👇
This is fixable though. We just have to give LLMs access to the docs. E.g. take whatever it is you want the LLM to have “fresh in memory because it just read it” (like we would do) and convert it to plain text (because this is what LLMs digest). @karpathy mentions this too.
To doo to option3️⃣👆here is the script github.com/FraPochetti/visua…. When I was working with o3-mini-high to incorporate Lambda functions backends into my Amplify web app, I just needed the LLM to be fully aware of THAT specific part of the docs, not the whole thing.
