I Got Tired of Using 6 AI Tools, So I Built One Instead

senthil kumar — Tue, 28 Jul 2026 05:10:30 +0000

I Got Tired of Switching Between AI Tools, So I Built One Platform

For every computer vision project, my workflow looked the same:

Search for a model.
Search for a dataset.
Annotate data.
Benchmark different models.
Train.
Deploy.

Every step required a different tool, different UI, and different workflow.

After doing this repeatedly, I started building MLForge to see if I could bring everything into one place.

Today it includes:

20,000+ AI models and datasets from major providers
Built-in annotation studio
Benchmark dashboard
Inference workspace
Computer vision training pipeline
Deployment interface

The project is completely free and designed to run locally on Windows, Linux, and macOS.

This is still an active project, and I'm continuously improving it based on feedback from other developers.

I'd really appreciate your thoughts:

What features would you expect from a platform like this?
What would stop you from using it?
What would make it part of your daily workflow?

Website:(https://mlforge.in)

Demo: [https://youtu.be/V0YBemyTkG8?si=CkOniaPa_Mwp7_B9]

Building a desktop-first Vision AI IDE — early thoughts & feedback welcome

senthil kumar — Thu, 01 Jan 2026 07:12:56 +0000

Post:

Hi DEV friends 👋,

Over the past month I’ve been tinkering with an idea: Ml Forge , a desktop-first IDE for Vision AI.

The motivation is simple: most Vision AI workflows feel fragmented. You jump between scripts, configs, dataset tools, training dashboards, and export utilities. My goal is to unify that into one place — with a no-code but still controllable approach.

Here’s what I’ve got so far:

Dataset manager (versioning, splits, validation)

Annotation studio

Training engine with live GPU metrics

Reproducible runs (dataset → config → model)

Logs & metrics

Inference & benchmarking

Export to ONNX, TensorRT, CoreML, TFLite, OpenVINO

Everything is UI-driven — no training scripts or deployment code.

I’d love to hear your perspective:

Does “no-code but controllable” resonate for Vision AI workflows?

What would make a tool like this useful in your day-to-day?

What blockers or concerns do you see?

I’m still early in development, so feedback from this community would be super valuable. Happy to dive into technical details or design trade-offs if anyone’s curious.

This tone works well on DEV because it feels like you’re sharing a journey and inviting collaboration, rather than promoting a product.

Would you like me to also add a short “storytelling hook” at the start (like how you struggled with fragmented workflows yourself) to make it more engaging for DEV readers?
refer the website to get info

DEV Community: senthil kumar

I Got Tired of Using 6 AI Tools, So I Built One Instead

Building a desktop-first Vision AI IDE — early thoughts & feedback welcome