![[mldp-newsletter]-sep-2023 — machine-learning-communities:-highlights-and-achievements](https://prodsens.live/wp-content/uploads/2023/10/15848-mldp-newsletter-sep-2023-machine-learning-communities-highlights-and-achievements-550x367.jpg)
[MLDP Newsletter] Sep 2023 — Machine Learning Communities: highlights and achievements
Let’s explore highlights and accomplishments of the vast Google Machine Learning communities over the month. We appreciate all the activities and commitment by the community members. Without further ado, here are the key highlights!
Keras Community Day finished!
With the contents from the Keras Sprint, the Keras Community Day campaign by the ML communities was done throughout July to September. More than 10,300 attendees around the world participated in local event!
In September, Keras Community Day in Bandung, Bangalore, Bangladesh, Bauchi Nigeria, Bhubaneswar (video), Bolivia (video), Boston, Chandigarh, Coimbatore, Colombia (video), Hajipur, Hong Kong, Islamabad, Jalandhar (video), Kuala Lumpur, Los Angeles, MLAct, Mysuru, Nashik, Nigeria, Nuremberg, Peru (video), Salem, Seattle (video), South Korea, Sri Lanka, Vietnam (video), Indore, Pune, Surat, and 13 other places were successfully hosted.
Generative AI/LLM
MakerSuite/PaLM Sprint
https://medium.com/media/7054899f1668558e7588848494f5549d/href
Building a Q&A Chatbot using Google PaLM 2 Model on Neo4j Database by ML GDE Bhavesh Bhatt (India) explains how to harness the power of PaLM 2 through Makersuite interface to create a chatbot that helps you ask questions to your Neo4j Database using natural language. He explains the basics of PaLM 2 and how it can be used to power chatbots. He also gave a talk on this topic at Keras Community Day Jalandhar. Another tutorial, Create a Smart Voice Assistant using Google’s PaLM 2, MakerSuite, Neo4j , Whisper, Python & Gradio provides a step-by-step guide to creating a voice-controlled assistant.
How to use Gen AI for Social Media Marketing with PaLM API and Maker Suite(Spanish version) by ML GDE Nathaly Alarcon Torrico (Bolivia) is a guide to create a website that generates marketing content using PaLM API, MakerSuite and Streamlit. She also crafted videos on the same topic and you can check here (English | Spanish).
https://medium.com/media/bb39b7e23d7fe47310b15bd56defb783/href
PaLM in Flowise (no-code langchain) by ML GDE Martin Andrews (Singapore) explains how the new PaLM API functionality he added to Flowise works in practice.
A Defacto Guide on Building Generative AI Apps with the Google PaLM API by ML GDE Victor Dibia (United States) dives deep on how to access/call the PaLM API (MakerSuite, Vertex AI client libs, REST api), implementing a useful task, and developer notes.
Creating a Q&A Bot Quickly in Google Colab with Makersuite and PaLM API by ML GDE Esther Irawati Setiawan (Indonesia) walks you through each step of the process, from designing the prompt to implementing the code in Colab. She shared another tutorial using Flutter for a mobile application: Building an AI Chatbot using Flutter with Makersuite and Palm API: A Step-by-Step Guide.
PaLM.NET. by Angular GDE Will Huang (Taiwan) is a PaLM API wrapper for .NET that supports .NET Standard 2.0.
LLM/Generative AI
AI Sommelier Built with PaLM API and LangChain by ML GDE Gabriel Cassimiro (Brazil) is about a project to build an LLM-powered application using PaLM API. For personalized wine recommendations, he built a solution using LangChain chains, Google Embeddings, Pinecone as the VectorDB, and Streamlit as the interface to interact with users.
Important update of Google Bard by ML GDE Yucheng Wang (China) introduced the latest updates of Bard and some other related products.
Getting Started with PaLM API in Vertex AI by ML GDE Olayinka Peter Oluwafemi (Nigeria) and VertexAI / GenAI Workshop : Getting Started by ML GDE Sara EL-ATEIF (Morocco) were both introduction sessions about PaLM API & VertexAI for beginners.
https://medium.com/media/ff31801b98fd646239b55abd3d4d9e68/href
ML GDE Sam Witteveen (Singapore) shared several videos in the LLM field: PaLM2 from Scratch, PaLM 2 Meets LangChain, Google’s Bard can now SEE and SPEAK 40 new languages! recorded thousands of views separately. You can check his recent videos on the hot topics in LLMs.
In Two Voice Devs episode 158 — Picture an Embedding If You Will, ML GDE Allen Firstenberg (United States) talked about his recent work with an image embedding model from Google, including incorporating it into LangChain JS. Allen and Mark shared what that process was like, what this new model lets us do, and how well it works. Plus, Allen created LangChain JS class to connect to multimodal embeddings models in VertexAI and to connect VertexAI Matching Engine vector store.
Discover the latest generation of conversational AI, ChatGPT and Google Bard in our interactive workshop by ML GDE Svetlana Meissner (Germany) shared how to utilize the latest technologies in SMEs. From conception to implementation, she showed how to create your conversational AI using tools like Bard.
Adventures with MakerSuite and PaLM 2 by TFUG Singapore focused on what’s new in ML products and Could Next. In one of the sessions, ML GDE Sam Witteveen showed how you can fine tune models for various tasks and uses using VertexAI and get MakerSuite to help in creating data for various Fine Tunes.
Introduction to Generative AI and Google PaLM (video) by GDG Chandigarh covered the different types of Gen AI models, and how they can be used to create new and innovative products and services. A live demo for PaLM2 was included as well.
In GDG Cloud Zürich Meetup #9 (September) by GDG Cloud Zürich, Googler Riccardo shared how to use Vertex AI to create stories, summarize them and create contextual images.
Keras
Supercharge KerasNLP Models with Wandb by Anshuman Mishra (India) is the 2nd edition of the KerasNLP tutorial contributed to Keras.io This notebook serves as a guide for ML practitioners who use KerasNLP and need experiment tracking tool Weights & Biases support in their workflows. Anshuman also gave talks on the tutorial at TFUG Mumbai and on Modular NLP Workflows with KerasNLP.
How to survive to Naked and Afraid using an Autoencoder by ML GDE Arnaldo Gualberto (Brazil) presented different types of autoencoders with Keras and TensorFlow for a large variety of applications like: dimensionality reduction, embeddings search, image generation, segmentation, and etc.
Building a Face Recognition Search Engine AI System with TF, Keras, and Flask by MLNomads talked about the challenge in designing a system for face recognition, how to leverage pre-trained models, and how to seamlessly integrate components into a cohesive system.
Review Google I/O, Insight into the Future of Technology by ML GDE Xiaoquan Kong (China) shared the ML product updated in the I/O and introduced new features of Keras.
#KCD ML and LLMs: Changing the Dynamics of Communication and Natural Language by ML GDE Lesly Zerna (Bolivia) was a talk about LLM and how KerasNLP and MakerSuite can help you to get into the NLP world.
#KCD Making Keras Models go brrr with XLA (slides) by ML GDE Sayak Paul (India) showed how to optimize Keras vision models with XLA and the kind of speedup one can expect.
#KCD Venturing into the world of GenAI with Keras and TensorFlow by ML GDE Nitin Tiwari (India) shared his insights on harnessing the power of Keras and TensorFlow to build Gen AI models for CV applications. His other talk, Unleashing creativity with Keras and TensorFlow shared how to utilize Keras and TensorFlow for the development of DCGAN models.
#KCD Abheesht Sharma (India) has been contributing to KerasNLP as an open source contributor. He also gave talks titled: KerasNLP: Efficient Finetuning of LLMs using LoRA in two Keras Community Day events.
#KCD Embracing the Keras Ecosystem by ML GDE Soumik Rakshit (India) & ML GDE Suvaditya Mukherjee (India) was about how you can adopt and integrate the Keras ecosystem with the larger ML ecosystems utilizing its backend-agnostic nature to the maximum.
#KCD Embracing Keras Ecosystem and Intro to TorchModuleWrapper by ML GDE Ayush Thakur (India) and ML GDE Soumik Rakshit (India) introduced the new features of Keras Core — the ability to train a PyTorch model with Keras Core. They demonstrated the use of TorchVision models and HuggingFace in the Keras workflow.
#KCD Compete in Kaggle using KerasNLP by ML GDE Thierno Ibrahima DIOP (Senegal) was about how to use the KerasNLP library for text classification using pre-trained models and finetunning.
#KCD What’s new in Keras Core by ML GDE Kuan Hoong (Malaysia) shared its multi backend support; Keras Core by ML GDE Martin Andrews (Singapore) introduced the new Keras feature’s benefits; and Keras 3 is all you need by ML GDE Aritra Roy Gosthipaty (India) & ML GDE Aakash Nain introduced what Keras 3 is and why/how to use it.
#KCD Keras NLP and LLMs with MakerSuite by ML GDE Esther Irawati Setiawan (Indonesia) discussed some challenges of building ML models for NLP and how to overcome them.
#KCD Getting started with Keras and Machine Learning by ML GDE Patricia O’Callaghan Olivo (UK) was an introductory workshop. ML GDE Sara EL-ATEIF (Morocco) also gave An Introduction to KerasCV.
#KCD AI on Edge with Keras by Aashi Dutt (India) was to share the implementations of Keras on Edge devices along with project showcases.
Kaggle
#KCD Usha Rengaraju (India) and Seshu Raja participated in Kaggle competition, LLM Science Exam: Use LLMs to answer difficult science questions and won silver medal.
And Usha gave talks on Compete in Kaggle using Keras Core and KerasNLP (notebook).
A New Way to Chat with Friends & Communities by ML GDE Sayak Paul (India) was an online event about DALL-E 2 & diffusion models. It was hosted by Kaggle.
Data analysis and machine learning in competitive data science ML GDE Luca Massaron (Italy) was a talk about Kaggle and how it can help develop your professional career in ML and AI.
On-device ML
On-device ML using MediaPipe by ML GDE Sachin Kumar (Qatar) was one of the I/O Extended sessions discussing ODML using MediaPipe. He showed how MediaPipe can be used to build applications that can recognize faces and track objects, hand gestures, poses and more that can be easily deployed to various devices.
Responsible AI
https://medium.com/media/8d60b522140a4c570e7ee1a97cfcc81b/href
Beyond the Buzzwords: Demystifying AI by ML GDE Xavier Portilla Edo (Spain) was a podcast session in Google I/O Connect Amsterdam. They discussed common misconceptions surrounding AI and tthe true nature of this transformative technology, exploring ethical considerations, privacy concerns, and the importance of responsible AI development.
ML Research
Paper reviews by ML GDE Grigory Sapunov (UK) for Chain-of-Thought Prompting Elicits Reasoning in Large Language Models, Uncovering mesa-optimization algorithms in Transformers, and Generative Agents: Interactive Simulacra of Human Behavior shared through his social channels.
Resume Shortlisting and Ranking with Transformers by ML GDE Akshay Kulkarni (India) helps the human resource domain eliminate the time-consuming recruitment process task such as reviewing resumes. Sentence-BERT (SBERT) network is a Siamese and triplet network-based variant of the BERT architecture, which may generate semantically significant sentence embeddings.
Product Specific Highlights
#VertexAI Custom model training & deployment on Google Cloud using Vertex AI in Go by ML GDE Paolo Galeone (Italy) introduces a more advanced usage allowing experienced ML practitioners to have full control on the pipeline from the model definition to the hardware to use for training and deploying.
#Cloud ML GDE Minori MATSUDA (Japan) presented a keynote session in Google Cloud Next 23 San Francisco.
[MLStory] A Guide to Using VertexAI and Google’s Embeddings for Generative Projects
#VertexAI Three ML Story articles by ML GDE Ashmi Banerjee (Germany) explain how to use and leverage VertexAI for your ML projects: A Guide to Using VertexAI and Google’s Embeddings for Generative Projects; Exploring the Future of Text Generation: A Deep Dive into Vertex AI using Jupyter Notebooks; and Leveraging VertexAI and “Clever” Prompts for Textual Data Insights.
[ML Story] Computer Vision made easy with Google Cloud Vision API
#Cloud Computer Vision made easy with Google Cloud Vision API by ML GDE Nitin Tiwari (India) is an ML Story article serving as a comprehensive initiation for beginners to embark on their ML journey with Google Cloud.
https://medium.com/media/f691a3c7bebf28a92ad89db69d9674bb/href
#Dialogflow Two Voice Devs episode 156 — Go with the Dialogflow CX Flow by ML GDE Allen Firstenberg (United States) and Xavier Portilla Edo (Spain) talked about the latest additions to Dialogflow CX. The new system functions make some of the processing you can do on inputs easier and faster!
#Kubeflow #VertexAI Kubeflow Pipelines et Vertex AI (slides) by ML GDE David Cardozo (Canada) introduced Kubeflow pipelines and ML toolkits in Kubernetes and CNCF Meetup in Quebec.
#BigQuery Useful Tips for Using BigQuery Without Getting Scared by ML GDE Jéssica Costa (Brazil) presents some features of BigQuery that allow you to optimize costs in the tool.
#TensorFlow TensorFlow User Group Ghaziabad Launch Event by TFUG Ghaziabad was to spread awareness of TensorFlow in the region. Aashi Dutt participated here as a speaker and talked about getting started with TF as well as how to contribute to TFUGs.
#TensorFlow Predicting survival on the Titanic using TensorFlow.js by GDG Barcelona was a part of workshop session to learn and build a model to predict the chances of survival of the passengers of the Titanic using real information given about their sex, age, cabin, port boat, etc.
#TensorFlow #Flutter Road To DevFestFlorida: WebML, Flutter, and LLM integration by GDG Central Florida & Space Coast explored ways to integrate LLM modes in web and mobile environment using TensorFlow.js, Flutter, PaLM2, and etc.
In Global Hispanic and Latino Developers Share How They Use Google Tools, ML GDE Juan Guillermo Gómez (Mexico) was introduced as an ML expert, and he shared his experience and tips for using Google tools.
🏅ML GDE Chansung Park (Korea) and 🏅ML GDE Rubens Zimbres (Brazil) were awarded as DevLibrary Top Contributor Award Winner! Check out their archives here: Chansung, Rubense.
[MLDP Newsletter] Sep 2023 — Machine Learning Communities: highlights and achievements was originally published in Google Developer Experts on Medium, where people are continuing the conversation by highlighting and responding to this story.
