alvisesus

Passionate Software Engineer, Author,
Eager to learn new things, Fell in love with Elixir.

[Paulo Valente @polvalente]

Very early stage, but here's NxEigen!

A new native Nx backend meant for lightweight memory footprints. This is an Arduino Uno Q with 300MB memory allocated *in total* (BEAM and Linux included)

https://t.co/ENPPI3WTKL

#MyElixirStatus

[alvisesus @alvisesus]

@MortenMLund Hey, just opened X after a while. I've answered you under the issue: https://t.co/cMdlwP5gm0

[alvisesus @alvisesus]

🚀 Just released YOLO v0.2.0 for Elixir!
Now with YOLOX support, custom-trained models, and faster postprocessing thanks to Nx.Defn.
Perfect for real-time object detection in Elixir.

https://t.co/6voKpWAjTv

#myelixirstatus

[Code BEAM @CodeBEAMio]

Give your Elixir apps vision! @alvisesus shows YOLO object detection integration into the BEAM - from custom training to real-time edge tracking at Code BEAM Europe 2025.
https://t.co/uS2jhO1W3b

[alvisesus @alvisesus]

Built a parking sensor for my impossibly tight garage 🚗

Problem: 4.7m car, 4.8m garage space 😅

Solution: ESP32 + ultrasonic sensor + LED strip that shows distance with colors

https://t.co/sDKjOIQTtV

#arduino #esp32 #electronics

[Code BEAM @CodeBEAMio]

🔥 CodeBEAM Europe lineup: @alvisesus (object detection), Arne Stoll & @anne_biene (GenServer TLA+), Brett Cameron (legacy porting), @elbrujohalcon (Elvis linter), Carsten Rösnick-Neugebauer (AI robotics)! https://t.co/5vuwwHTR61

[Elixir Radar @elixir_radar]

Elixir Radar issue 478 is out! 📣

You can read it here: https://t.co/Qgh8J1RUCF

This issue comes with content from @mpraski96 @ZachSDaniel1 @alvisesus @matthewsinclair , Thibaut Decaudain and Ettore Berardi. Thank you!

#ElixirLang #MyElixirStatus

[alvisesus @alvisesus]

Fine-tuning YOLO models for custom object detection 🤖

See how to transform a generic 80-class detector into a specialized system and integrate it with Elixir applications. Applicable to sports analytics, manufacturing QC, and more!

https://t.co/UTrnVz3kF3

#myelixirstatus

[Elixir Radar @elixir_radar]

Elixir Radar issue 477 is out! 📣

You can read it here: https://t.co/mhkMpc0zsc

This issue comes with content from @codestirring @alvisesus @katafrakt_pl @ThatArrowsmith @kraleppa and Fredrik Teschke. Thank you!

#ElixirLang #MyElixirStatus

[alvisesus @alvisesus]

@antirez @lcamtuf the longest path could not be the "best" path/life.

[alvisesus @alvisesus]

@MortenMLund yes it is! You can fine-tune the ultralytics models and YoloX as well. At the moment the Elixir library doesn't support custom models, but it will very soon!

[alvisesus @alvisesus]

🚀 Just released a video tutorial on my Elixir YOLO library for real-time object detection! Check out what you can build with just a few lines of code 👇
#myelixirstatus

https://t.co/9jrHNGKPlT

[Elixir Radar @elixir_radar]

Elixir Radar issue 457 is out! 📣

You can read it here: https://t.co/9vkWnYbcQx

This issue comes with content from @katafrakt_pl @JohnElmLabs @M3gy0 @alvisesus , Dmytro Kozin and Andrew Berrien. Thank you!

#ElixirLang #MyElixirStatus

[alvisesus @alvisesus]

@NoblNjord evision https://t.co/fq1rwpyisU to read frames from the camera, scale them and convert them to a tensor.

[Elixir Radar @elixir_radar]

Elixir Radar issue 455 is out! 📣

You can read it here: https://t.co/m6hFvGBkRm

This issue comes with content from @Mike_Andr @ananthakumaran @sasajuric @alvisesus @timgremore and Chris Houhoulis. Thank you!

#ElixirLang #MyElixirStatus

[alvisesus @alvisesus]

Just thinking: Calling most "open source" LLMs (like LLaMA) feels off. Sharing weights & architecture without training data, hyperparams etc.. is like sharing a compiled binary & calling it open source. The training process is the real source. Without it, it’s just a black box.

[alvisesus @alvisesus]

🎉 Super excited to share my first Elixir library: YOLO! 🚀

Runs YoloV8n in 38ms on my mba M3 using YoloFastNMS

https://t.co/BqBWV55ccm

#myelixirstatus

[alvisesus @alvisesus]

5/ VisionOS Experiments

Exploring VisionOS by anchoring virtual objects to real-world items (like my car engine bay)

Post on reddit with full video: https://t.co/JwDfPjDVQo

[alvisesus @alvisesus]

It’s been a while! After focusing on ML and financial data full-time, I’m now consulting and lately I’ve been diving into:

- Video sharing with LiveView + FLAME
- YOLO object detection in Elixir
- Race car telemetry with Nerves
- VisionOS experiments

Here's a quick update👇

[alvisesus @alvisesus]

4/Race Car Telemetry with Nerves + wireless sensors

As a racing enthusiast, I’m building a telemetry hub with Nerves + ESP32 boards, to save real-time data of tire temps (thermal cameras), RPM, gear, oil temp (CAN-BUS), GPS + accelerometer.