A revolutionizing empathic AI solution built by an Intel® Student Ambassador using Intel® Tiber™ Developer Cloud
During critical emergency situations such as natural calamities, the emergency services like 911 receive hundreds and thousands of requests for urgent assistance. A shortage of call centers' staff at such times may hinder the ability to mitigate critical emergencies. An Intel® Student Ambassador developed an empathic Generative AI (GenAI) solution called Dispatch AI leveraging Intel’s latest accelerated hardware, AI tools and frameworks on Intel® Tiber™ Developer Cloud platform to revolutionize emergency responses, providing emotion-aware assistance to both the calls and the dispatchers of the emergency call services.
The Dispatch AI project won multiple prizes at the University of California, Berkeley’s AI Hackathon held on June 22-23, 2024.
About the Intel® Student Ambassador
Bill Zhang is currently pursuing a Masters in Computer Science degree with a focus on AI at the University of Southern California. His journey as an Intel Student Ambassador showcases the power of combining innovative ideas with Intel's advanced cloud computing and AI technologies to create solutions that can potentially save lives and revolutionize industries.
About the Dispatch AI Project
In just 27 hours during the UC Berkeley's AI hackathon, Bill and his team integrated AI agents, LLM evaluations, and physical phone calling into a comprehensive SaaS platform. Dispatch AI processes emergency calls through Twilio*, utilizing a custom Mistral AI LLM fine-tuned on Intel Tiber Developer Cloud to generate responses. Simultaneously, the system employs OpenAI* GPT-4 to extract critical information from call transcripts. The team also incorporated emotion detection feature into the solution for more empathetic call responses.
The interactive UI of Dispatch AI looks as follows:
Fig.1: User interface of Dispatch AI
Architecture of Dispatch AI
Following is a comprehensive architecture design of Dispatch AI that helps visualize the communication flow across various components of the hardware and software stack used in the project:
Fig.2: Communication flow across Dispatch AI’s architecture
Intel’s AI Tools and Technologies Used
Bill and his team leveraged comprehensive Jupyter Notebooks readily available on Intel Tiber Developer Cloud throughout the development of the project. They also used Intel® Extension for PyTorch* for model optimization.
Results Achieved Using Intel’s AI Developer Resources
The resources utilized on Intel Tiber Developer Cloud helped achieve a remarkable reduction in inference time from 2 minutes 53 seconds to less than 10 seconds. Nearly 80% reduction in the processing time was attained using Intel’s powerful and optimized AI software.[1]
Impact and Recognition
Bill and his team won the first-place prize in the ‘Best Use of Intel AI’ track and a grand prize in the ‘Best AI for Good’ category at the hackathon, securing over $50,000 in prizes and investments. Each of the team members was also awarded a Lenovo* AI PC by team Intel. This achievement highlights the potential of Intel Tiber Developer Cloud in developing cutting-edge AI solutions for critical real-world problems.
Scope for Future Enhancement of Dispatch AI
The success of Dispatch AI has garnered interest from police departments around California, prompting Bill and his team to consider turning their project into a startup. If they proceed with the startup, they are planning to fine-tune the model on Intel® Gaudi® 2 AI accelerator GPU, further optimizing the solution for speed and latency.
Check out the complete project details of Dispatch AI here. |
What’s Next?
Sign up to Intel Tiber Developer Cloud today and get started with accelerated development of AI, HPC and edge-computing solutions! Explore the Intel Student Ambassador Program to utilize Intel® oneAPI tools and the Intel Tiber Developer Cloud in collaboration with next gen developers.
We encourage you to check out the AI, HPC, and Rendering tools in Intel’s oneAPI-powered software portfolio.
Useful Resources
- Intel Student Ambassador Program
- oneAPI programming model
- AI Developement Resources
- AI Tools from Intel
- AI Frameworks Optimizations
- Generative AI (GenAI) Resources
- Intel's Ecosystem Developer Hub
- AI/ML Ecosystem Developer Resources
Notices and Disclaimers
[1] Inference was performed on an Intel’s VM, specifically a 4th Gen Intel® Xeon® Scalable Processor with 56 cores, 512GB memory, and 2TB disk.
Performance varies by use, configuration, and other factors. Learn more at www.Intel.com/PerformanceIndex .Your costs and results may vary.
Intel technologies may require enabled hardware, software, or service activation.
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries.
*Other names and brands may be claimed as the property of others.
Technical Software Product Marketing Engineer, Intel