Ramya Ravi, AI/ML Software Marketing Engineer, Intel
Data Scientists and Developers are always looking for ways to boost their AI applications/solutions performance. Intel provides various AI tools and frameworks to accelerate end-to-end data science and analytics pipelines. Using optimized Intel oneAPI and AI tools allows users to build and optimize oneAPI multiarchitecture applications on Intel CPUs and GPUs.
This article puts emphasis on three on-demand tech talks that were delivered at oneAPI DevSummit for AI and HPC 2022:
- Using oneAPI to Predict Anonymous Web Visitor Behavior
- Hacking the Hackathon using FastAI and Intel® Extension for PyTorch*
- Spatial Single Cell Analysis Using oneAPI AI Analytics Toolkit
The speakers of these tech talks focus on how they built and optimized their applications using Intel oneAPI, AI tools, and, frameworks.
Tech Talk 1: Using oneAPI to Predict Anonymous Web Visitor Behavior
In this tech talk, Peter Ma, Co-Founder of SiteMana, explained how SiteMana Inc. benefits from oneAPI and daal4py. Daal4py is an easy-to-use Python* API that provides superior performance for your machine learning algorithms and frameworks. This is included in the Intel® Distribution for Python* as part of the Intel® AI Analytics Toolkit. SiteMana is an AI company that predicts anonymous visitor purchasing intent. Peter started off his talk by explaining the need to identify and retarget anonymous site visitors with high purchasing intent. The steps involved in the project are:
- AI prediction - Predict the likelihood of the visitor who is going to purchase.
- Real Time Communication – RT Communication with the visitors by popping up coupons instead of asking for sign-up via email.
- Unmask visitor email - Aggregated targeting of high intent purchasers.
Next, he presented a short demo of the overall project and highlighted that this project is powered by oneAPI (using daal4py and scikit-learn). Finally, he brought up that the users of his project focused on return on investment and retargeting higher purchasing intent visitors.
Watch the full video recording here.
Tech Talk 2: Hacking the Hackathon using FastAI and Intel® Extension for PyTorch*
This tech talk is about how Ankur Singh and Sai Rama Raju Penmatsa, Graduate Students from San José State University, aced the Hackathon at Intel Innovation 2022 using FastAI and Intel® Extension for PyTorch*. They started the presentation by talking about their Hackathon experience. Next, Ankur briefly explained about the problem statement from both business and deep learning perspectives.
- Train a Deep Learning model to separate weeds from plants.
- Target pesticide attack on weeds.
- The deep learning model implemented will be deployed on a drone. It must be computationally cheap and should have fast inference speed.
- Image classification problem - CNN (Convolutional Neural Network) model and Binary classification (Two classes - weeds and plants)
- Loss - Binary Cross Entropy
- Metric - Accuracy Score
Finally, he highlighted how they conducted various experiments and approaches. Check out the full video recording and solution for the hackathon to learn about their approaches in detail.
Tech Talk 3: Spatial Single Cell Analysis Using oneAPI AI Analytics Toolkit
In this tech talk, Abhishek Nandy, Co-Founder of Dynopii, kicked off his presentation by explaining important terminologies in spatial single cell analysis:
- Transcriptional Profiling - Process to identify a rare disease condition, ways to diagnose and check whether the body responds to treatment.
- RNA Sequencing - A sequencing technique which uses next generation sequencing to reveal the quality of RNA in biological sample at given moment and analyzing the cellular structure continuously.
- Gene Mutation - A change in a gene structure. Some mutations can lead to genetic disorders or illness.
- Single-Cell RNA - Single Cell RNA-Seq provides transcriptional profiling of thousands of individual cells. It allows us to study gene expression and gene mutation in rare diseases.
Squidpy, Intel® AI Analytics Toolkit (AI Kit) and, Intel® Developer Cloud are the main technologies used in this project. Next, he explained why to use Squidpy. Squidpy is a tool for the analysis and visualization of spatial molecular data. This tool will be ported to AI Kit and ported within Intel® Developer Cloud. Finally, he explained a case study using Visium Dataset. The following steps are performed:
- Load the huge datasets. From these datasets, spatial graphs are created, and image analysis is performed.
- Extract the image and summary features.
- Generate cluster annotation from the image features.
- Compute neighborhood enrichment to identify spot (node) clusters that share a common neighborhood structure across the tissue. This also helps to uncover molecular mechanisms associated with cell differentiation and disease progression.
This project is useful to study rare diseases, ligand receptors, and structural anomalies in single cell RNA. Watch the full video recording here.
Download and try the Intel® AI Analytics Toolkit (AI Kit) for yourself.
We encourage you to learn more about and incorporate Intel’s other AI/ML Framework optimizations and end-to-end portfolio of tools into your AI workflow. Also, visit AI & ML page covering Intel’s AI software development resources for preparing, building, deploying, and scaling your AI solutions.
About our speakers
Peter Ma, SiteMana Inc. Co-Founder
Peter Ma is Co-Founder of SiteMana, an AI company that predicts anonymous visitor purchasing intent. He is also an Intel Software Innovator, TED speaker and, Techstars Alumni.
Sai Rama Raju Penmatsa , Graduate Student at SJSU
Sai Rama Raju Penmatsa is currently pursuing his Masters in Software Engineering with Data Science as a specialization at San José State University. He is a Graduate Research Assistant in Computer Vision.
Ankur Singh, Graduate Student at SJSU
Ankur started his own company in India after completing his undergraduate program. The company is called “AI Adventures,” which provided AI/ML solutions to businesses. After 3 years, he joined Zoop.one as the ML team lead. Currently, he is pursuing his Masters in Software Engineering at San José State University.
Abhishek Nandy, Dynopii Co-Founder
Abhishek Nandy is Co-Founder of Dynopii. He is an Intel Black Belt Developer and has presented his research work on Reinforcement Learning at ACM SIGGRAPH 2018. He is an invited educator at several leading premier education institutes in India.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.