Comet.ml: supercharging your machine learning workflow
I’m very excited to announce that I’ve joined Comet.ml as Product Lead!
After graduation, I became a product manager for Watson Studio, a powerful end to end machine learning platform. This role was an amazing opportunity to learn all about enterprise product management and ramp up on machine learning. Everyday, my goal as a product manager was to dive into our users’ workflows and see how we could make this complex vision of making sense of data seamless and effective.
One common theme came up again and again after working with countless data scientists and machine learning engineers: how can you enable us to build things in-house (so we can take advantage of emerging open-source developments) in a scalable, collaborative way to prove ROI for our team?
These major pain points around managing machine learning workflows explode when: (1) teams expand, (2) choice/use of machine learning frameworks expand, and (3) both data architectures and the data itself shifts.
Many teams were leaning on Github to alleviate some of these pain points. But Github still has not adapted for machine learning needs — enter Comet.ml.
Comet.ml helps tracks meaningful information regarding the environment such as your code, dependencies, hyperparameters, and metrics.
Because of Comet’s lightweight approach, the product works regardless of which frameworks and libraries you use to build your models. There are many other benefits:
- being able to train your models on your own choice of infrastructure (yes, this includes your 2015 MacBook 💻)
- faster experimentation by having your models performance metrics sit closer to your data + code
- Comet.ml’s hyperparameter optimization feature
The next weeks, months, and years(!) will be devoted to building out and growing this awesome product!
If you’re a data scientist/machine learning engineer who thinks Comet.ml can help your team, send me a message at firstname.lastname@example.org 😃
Read about Comet.ml: