Sign In

Chris Chang

Portfolio Manager at Elemental Excelerator

Chris Chang is the Director of Portfolio Experience at Elemental Excelerator, a role in which he leads the Strategy Track and supports diverse pre-seed to Series A stage startups.1 His responsibilities include:

  1. Facilitating interactions to accelerate important outcomes for portfolio companies
  2. Driving operations with the Elemental team, consultants, clients, partners, and cohort members
  3. Supporting the commercial growth of the portfolio
  4. Delivering a world-class accelerator program

Professional Background

Prior to joining Elemental Excelerator, Chris spent 7 years at Techstars, a global accelerator company.1 During his time there, he:

  • Developed startup communities
  • Created accelerator programming
  • Worked on pilot products
  • Supported accelerator expansion to Australia
  • Managed pilot initiatives with large corporate partners, including Target Corporation, Metro AG, and the US Department of Defense

Education

Chris holds two degrees from the University of San Francisco1:

  1. Bachelor of Science in Business Administration
  2. Master's degree in Entrepreneurship and Management

Personal Interests

Outside of work, Chris enjoys:

  • Ocean activities
  • Training for various sports competitions
  • Experimenting in the kitchen
  • Dreaming about surfing destinations

Chris Chang's LinkedIn profile can be found under the username chris96825.2

Highlights

Apr 26 · linkedin

I'm excited to share our applied research team has just released the first long context length adapted LLama3 LLM (260k token context length), in partnership with Crusoe.

It's our first showcase that SOTA LLMs can learn to operate on long context with minimal training (< 200M tokens, powered by Crusoe's compute) by appropriately adjusting RoPE theta. And it was trained on Crusoe’s latest L40S cluster.

Why do long context lengths matter for enterprise?

Imagine you're trying to use AI to interpret an email while using an MSA or privacy policy as reference (these things are long!). Your only choice so far has been to chunk these documents into RAG and pray that you're retrieving the right context.

However, documents aren't perfect glossaries for reference - often context is mixed and blended across the entire document.

With long context length models, this problem is solved - you're able to finally leverage all of the relevant private data without worrying about losing any critical information.

This was a non trivial amount of work by our research team as there are no commercially available training and inference frameworks that work for this kind of development. Huge props to them, and it's an indicator of things to come.

Check out our model on Hugging Face: https://lnkd.in/gCxhprr5

#Gradient #GradientAI #LLM #Llama3 #Crusoe #ContextLength

Related Questions

What are Chris Chang's main responsibilities at Elemental Excelerator?
How did Chris Chang contribute to Techstars before joining Elemental Excelerator?
What specific programs does Chris Chang lead at Elemental Excelerator?
What is Chris Chang's educational background?
How does Chris Chang support the commercial growth of startups at Elemental Excelerator?
Chris Chang
Chris Chang, photo 1
Chris Chang, photo 2
Add to my network

Location

San Francisco Bay Area