MLX India Community Meetup #1

"How many of you used MLX for the first time?" - half the hands in the room went up.

We are MLX India, a community of builders working with MLX, local LLMs, and AI on Apple Silicon. Kautuk and I (Sabesh) are co-organizers of the community. We have over 300 builders on our WhatsApp channel, where we share what we're building and interesting finds on a rolling basis. Members come from a wide range of teams - Apple, Swiggy, PhonePe, Raycast, LM Studio, and Hugging Face, to name a few.

On May 3rd, 2026, we hosted our first MLX India Community Meetup, and the outcome was strong. We brought together Swift and iOS developers, builders working with local LLMs, and AI researchers under a single roof - and roughly half the attendees were using MLX for the first time. Here's how the day went.

Talks

We opened with a lineup of four talks:

1. The MLX Framework 101 - Sabesh Bharathi
2. MLX Swift: Using MLX in iOS apps - Raj Raval
3. Running sustained LLM load on iOS: performance gotchas and navigating around constraints - Sahil AK
4. Boosting local model performance with post-training optimization (speculative decoding in MLX using DFlash) - Sabesh Bharathi

Q&A was active throughout, and the discussion stayed sharp across all four sessions.

Build Session

After the talks, we moved into a build session. The brief was open: build something with MLX - an iOS or macOS app, a Python project using mlx-lm, or research work like benchmarking and model quantization.

The number of first-time MLX users in the room was the real standout. They got hands-on quickly, helped along by experts already working in the space. App developers guided researchers on building proper iOS apps in Swift and adhering to HIG. Researchers, in turn, helped app developers pick the right models and apply the right techniques to squeeze performance out of Apple Silicon. That cross-pollination was exactly what we'd hoped to see.

One highlight worth calling out: an attendee on a Windows machine spun up an Amazon EC2 instance with Apple M2 and Metal - just so he could build his MLX-based app, which used mlx-lm, mlx-vlm, and mlx-audio under the hood. That kind of commitment was visible across the room.

Demos

After four hours of building, the top projects were demoed to the rest of the room. The quality was genuinely impressive - iOS apps, macOS apps, custom models trained and fine-tuned with MLX on Apple Silicon, MLX-based TTS and STT pipelines, and apps making use of tool use and structured outputs.

Wrap

The day ended on those demos. Everyone walked away with something new - a technique, a contact, or a project worth shipping. We believe the future of AI is local, and on the Apple platforms, that means MLX. The builders in the room helped shape a small piece of that future.

You can download pictures captured and edited at the event, here!