Welcome to our recap of our inaugural Intel® Liftoff Days 2024, a unique virtual hackathon event tailored exclusively for the program members. This week-long immersive tech acceleration experience offered our startup community a distinct opportunity to explore various cutting-edge topics with Intel.
From LLM and RAG pipelines to generative AI and language modeling, our hackathon pushed the boundaries of innovation. Our event was all about fostering collaboration and new ideas, and we saw some truly inventive and original projects being presented.
Our mentoring team held one-on-one sessions with our members throughout the week to provide hands-on guidance and practical advice they could apply to their startups.
Our Goals for Intel® Liftoff Days in June 2024
Members received access to exclusive workshops with Intel’s top AI experts to elevate their startups’ tech stack. They got hands-on experience enhancing code and AI performance on the Intel® Tiber Developer Cloud, using the latest Intel hardware.
Startup founders had the opportunity to present AI-powered projects to Intel, acquiring unique co-marketing opportunities and engaging with top minds at Intel and peers within the Intel® Liftoff community.
Participants who were nominated on our Demo Day, where projects were presented, also secured the opportunity to have extended cloud privileges.
What We Covered
Our workshops and ideation sessions covered a broad range of AI-related topics, each designed to provide members with cutting-edge techniques and inspire new conversations and ideas.
Here’s what we spoke about in our sessions during the week:
Monday
We discussed:
- An introduction to Intel® Tiber Developer Cloud
- Stable diffusion for synthetic data generation
- Setting up Intel® Tiber Developer Cloud for running Stable Diffusion and LLM
- LLM Zero shot finetuning.
Tuesday
Our sessions covered:
- Gemma model finetuning using LoRA on intel GPUs.
Wednesday
We introduced members to:
- Techniques for LLM quantization
- The Whisper Speech-to-Text integration.
Friday
Projects were pitched on our Demo Day by 12 AI startups at the forefront of their game. Each startup had three minutes to pitch their ideas, including:
- The starting point for their project
- The stage of their project
- A definition of their project
- Their learnings and next steps.
We also spent time recapping the event and asking members for feedback. Members reflected on what they had learned and said they appreciated our mentors' willingness and openness to provide guidance and feedback on their projects.
The Industries We Heard From
Our event was not limited to a single industry. We heard from startups that spanned a diverse range of industries, including legal, energy, medical, analytics, and even name-entry recognition.
Other industries we heard from included computer vision, LLM copilots, genAI power image generation, homographic encryption, retail product recommendation, drone path planning and obstacle avoidance.
The Techniques We Observed
Throughout our workshops and mentoring sessions, we covered various techniques, including LLMs, RAG, finetuning, and time series prediction using LLMs.
Projects We Loved
One of the first projects presented was by a startup called Pixel ML. Their goal at our hackathon was to optimize their Agentic AI workflows on Intel® Tiber Developer Cloud, enhancing their capacity to generate personalized content at scale.
Their project, titled AgenticFlow AI is an AI-native platform that automates the entire content marketing supply chain, from ideation to distribution, using no-code workflows powered by intelligent AI agents.
Another project we really enjoyed was presented by LastMinute. The project included AI-driven shift matching and predictive analytics leveraging Intel® Tiber Developer Cloud. Their goal at our hackathon was to develop a prototype that integrates machine learning models to predict staffing needs based on historical data and current trends.
The projects that were presented spanned across Intel CPU, GPU, and Habana Gaudi’s. We’ll be covering all the projects presented at our event in more detail, so keep an eye out for those articles. We also turned our attention to creating cost effective AI solutions.
Startups Gallery
Aleph Innovations has developed a suite of services that centralize, automate, and improve scheduling and referral processes to help improve the lives of people everywhere. | CleanML is a SaaS application for enterprise ML teams to efficiently identify optimal models. |
CloudConstable delivers animated AI powered personalities with purpose – for Retail, Hospitality and Aviation. | Bezoku is delivering small, dialect focused language models for powerful use cases including text generation, customer service and sentiment analysis. |
kAI is your intelligent personal assistant and AI-powered task manager. kAI turns your vision of an effortlessly organized life into reality. | LastMinute has developed a solution to automate healthcare workforce optimization with AI. |
At Legal-Pythia, we're at the forefront of legal technology, revolutionizing the world of legal documentation, using Explainable AI. | Maplewell Energy modernizes energy demand with an AI-driven platform for distributed energy storage. |
Pixel ML is a go-to-market AI suite. It uses AI-powered agents to 10x your company growth. | The Mango Jelly is an AI copilot and collaborative workspace for marketing team using Generative AI. |
Venturely is the all-in-one AI- & data-driven platform for organizations of all sizes to collaborate, learn and innovate. | VOTIX is a powerful and complete cloud-based SaaS solution designed to support all drone manufacturers automating drone-based workflows, powered by real-time AI: |
What we learnt
Key takeaways include the power of teamwork in integrating diverse AI techniques, using advanced data security like Tensor Seal and homomorphic encryption, creating budget-friendly AI by combining open-source models, and applying large language models and deep learning to improve predictions in energy demand and staffing:
- Collaborative Problem Solving and Integration: The hackathon showed how important teamwork is for solving tough AI challenges. Diverse teams worked on projects like video generation, marketing analysis, and EHRs, demonstrating how combining different AI techniques can lead to cool solutions.
- Data Security and Privacy: Keeping data safe was a big focus. Teams looked into encrypting data with Tensor Seal, and Aleph Innovations explored homomorphic encryption to keep client data private in machine learning.
- Cost-Effective AI Solutions: We discussed how to make AI solutions budget-friendly. Teams used a mix of bigger open-source LLM models from hosted APIs and smaller CPU models. They, along with Last Minute, took advantage of open-source models.
Innovative Use of LLMs and Time Series Forecasting: The hackathon highlighted creative uses of large language models and deep learning. Maplewell Energy boosted predictive analytics for energy demand, and LastMinute worked on predicting staffing needs.
What we achieved
Hackathon participants gained deeper insights into various AI techniques, such as LLM fine-tuning, time series forecasting with Chronos LLM models, and the use of homomorphic encryption, enriching their technical expertise.
Several teams successfully developed prototypes during the hackathon. These included drone operations for public safety, customer service AI, and developed solutions for predictive staffing needs, demonstrating practical applications of AI technologies.
Participated startups were also above to build on their expertise in specialized areas such as improved query processing and demonstrated the development of efficient query engines using recursive indexing and structured data management to improve query accuracy and retrieval speed.
Votix showcased the power of Real-Time prompt processing and the potential of AI in enhancing drone operations for public safety, illustrating how real-time prompts and optimized routing can improve emergency response efficiency and accuracy.
"Our recent participation in the Intel Liftoff Hackathon has profoundly transformed our approach to innovation in the drone industry. The event was a game-changer, significantly enhancing our understanding of integrating LLM and Visual Analytics to push the boundaries of what's possible. A big shoutout to the Intel Liftoff team! Their unwavering support and deep expertise were crucial in helping us unlock the full potential of Intel's technology. We're thrilled about the upcoming features on our roadmap that will leverage these advancements, bringing immense value to our customers,"
- Edwin Sanchez, CEO, VOTIX.
Most importantly, the focus was on scalable AI solutions. Maplewell Energy utilized Intel Liftoff member's Prediction Guard for secure, open-source-compatible API usage, setting the stage for wider adoption and implementation across different industries.
“With the team support at Intel Liftoff Days, we felt comfortable pushing our boundaries and trying something risky. Without this program, we would have not explored LLMs in our application,”
- Matthew Irvin, CEO, Maplewell Energy.
Our Next Adventure with Intel Liftoff Days
The Intel® Liftoff Days Q2 was a massive success, and our next hackathon, held from July 29 – August 2, 2024, continued the momentum. Stay tuned for more exciting events and insights in the future.
Are you an early-stage AI startup looking to make your mark? Apply now to join the Intel® Liftoff Program for Startups and unlock a world of possibilities. This is your chance to collaborate with industry-leading innovators, supercharge your projects, and propel your start-up to the next level with Intel’s support.
I'm a proud team member of the Intel® Liftoff for Startups, an innovative, free virtual program dedicated to accelerating the growth of early-stage AI startups.