Artificial Intelligence (AI)
Discuss current events in AI and technological innovations with Intel® employees
544 Discussions

Intel sponsored UC Berkeley’s AI Hackathon

Nikita_Shiledarbaxi
0 0 1,131

Developers leveraged AI PCs and Intel® Gaudi® AI Accelerator on Intel® Tiber™ Developer Cloud

 

University of California, Berkeley recently organized an AI hackathon on June 22-23, 2024. Intel, being an official cloud compute sponsor of the event, provided the attendees with free access to the Intel® Tiber™ Developer Cloud including Intel® Gaudi® AI Accelerator. For an even more exciting experience, team Intel also provided the developers with an amazing opportunity to learn about and try their hands on Intel® Core™ Ultra Processors (or Meteor Lake) based AI PCs! Moreover, the attendees learned about Intel’s resources for AI/ML and GenAI, AI frameworks optimizations, AI development software portfolio, Intel’s ‘AI Everywhere’ campaign, Intel® Liftoff Program for AI startups, Intel® Student Ambassador Program and Educator Program.

 

Highlights of Intel’s presence at the hackathon

Here’s a recap of the UC Berkeley’s AI hackathon:

  • 1300+ students attended the event.

  • 475 new users signed up for Intel Tiber Developer Cloud to leverage Intel’s latest accelerated hardware and software optimizations.

  • 225 students learned about various Intel® AI technologies through 3 workshops, and interactive sessions about AI PCs and startups upliftment conducted by team Intel.

  • 55 AI PC development kits were provided for developers to have hands-on experience at the Intel AI PC Lab set up at the event.

  • 75 students grabbed an opportunity to learn about each of the Intel® Gaudi®2 AI Accelerator and AI PCs.

  • The Intel Liftoff team collaborated with the Berkeley SkyDeck team to organize a meetup where students having a startup or interested in starting one were guided on how Intel’s technology can assist them advance in their entrepreneurial journey.

 

Winner Projects of the ‘Best Use of Intel AI’ Track

Following were the top 3 projects that best utilized Intel’s AI technologies at the hackathon:

  1. Dispatch AI, the 1st-prize winner, is an empathic AI agent eliminating 911 wait times during critical emergencies. The team potentially leveraged Intel Tiber Developer Cloud including various Jupyter Notebook available on the platform. It extensively utilized Intel’s AI tools, particularly Intel® Extension for PyTorch* for model optimization. It also implemented a fine-tuned Mistral LLM for specialized emergency response using Intel Tiber Developer Cloud. Two of the team members are part of the Intel Student Ambassador Program.

  2. Batteries by LLM, the 2nd-prize winner, employs LLMs to predict the structure of new lithium-ion electrolytes from text description and optimizes it further using first principles modeling to fight climate change. Using Gaudi cards on Intel Tiber Developer Cloud, the team fine-tuned a Llama-2-7B model that converts natural text to modeling input files that reflect atomic positions.

  3. Accel, the 3rd-prize winner, is an empathic AI tool that acts as a chemistry tutor. The team leveraged Intel® AI PC and Intel Gaudi2 accelerator to develop the core capabilities of the project. It distilled a model using Intel Gaudi2 by fine-tuning synthetic data from Llama 70B to 3B model.

The winners were awarded free cloud credits for further enhancement of their hackathon projects on Intel Tiber Developer Cloud. Moreover, each member of the 1st-place winning team also received a Lenovo* AI PC. The 2nd -place winners each received a Sparkle Intel® Arc™ A770 GPU Titan OC Edition and the 3rd-place winners each won a Sparkle Intel® Arc™ A750 GPU Titan OC Edition.

 

Projects Harnessing the Power of AI PCs

Around 50 students tested out and learned about the exceptional capabilities of AI PCs at the event. Some of the hackathon projects listed below also utilized the power of AI PCs:

  1. Accel (as mentioned above) won the 3rd-prize in the ‘Best Use of Intel AI’ track.

  2. ASL Bridgify, a winner of the ‘Best AI for Good’ track at the hackathon, is an educational platform for learning American Sign Language (ASL). The team used Intel Extension for PyTorch and Intel® Extension for TensorFlow* on CPU and GPU environments available on Intel Tiber Developer Cloud. One of the team members is an Intel Student Ambassador.

  3. Math Made Easy employs optimized LLM for providing comprehensive step-by-step solutions to mathematical and logical reasoning problems.

  4. Interview Aid is a technical interviews simulator platform powered by AI.

  5. DPAncestory is a platform that processes genomic data using local differential privacy algorithms, maintaining individual privacy.

  6. StudyBuddy is an empathic chatbot governed by teachers that acts as an interactive AI study partner for students.

  7.  Roadie is an AI-powered solution for a seamless driving experience with facial recognition, audio chatbot, and potholes/road signs detection features.

 

Insightful Workshops Conducted by Team Intel

Apart from Intel Liftoff – Berkeley SkyDeck meetup for startups and Intel AI PC lab, team Intel organized the following three interactive workshops to educate the hackathon attendees about Intel’s latest technologies.

  1. Workshop #1: A kickoff session where 75 attendees explored Intel Tiber Developer Cloud, were onboarded on the cloud platform for developing their hackathon projects and learned about the possibilities of utilizing Intel-provided AI PCs at the event.

  2. Workshop #2: The attendees learned how to fine-tune a 7B Llama model in less than 6 minutes on an Intel Gaudi2 processor and accelerate training and inference across 8 Gaudi2 cards using tools like DeepSpeed*. Interested students then obtained Gaudi cards for their hackathon project.

  3. Workshop #3: The attendees learned how to build their own chatbot on the 4th Gen Intel® Xeon® Scalable Processors (or Sapphire Rapids) featuring the Intel ®Advanced Matrix Extensions (Intel® AMX) instruction set used to accelerate deep learning workloads. Students walked away with the knowledge to run LLM model inference on Xeon processors.

 

What’s Next?

Dive deeper into Intel® AI PCs and leverage their astounding AI acceleration capabilities through AI PC Development resources including devkits.

Get started with Intel Tiber Developer Cloud for accelerated AI development. If you are new to the cloud platform, sign up today! Also explore the Intel® Gaudi® AI accelerator and expedite high performance LLM and GenAI applications.

Join the Intel Liftoff Program to advance your early-stage AI startup into the venture capital industry. To collaborate with developer communities for innovating with Intel® oneAPI unified programming model and Intel Tiber Developer Cloud, explore the Intel Student Ambassador Program. For software educators willing to integrate industry-standard heterogeneous computing resources into institutions’ STEM courses, check out our Educator Program.

We also encourage you to learn about AI, HPC, and Rendering tools in Intel’s oneAPI-powered software portfolio.

 

Additional Resources      

 

About the Author
Technical Software Product Marketing Engineer, Intel