[MLDP Newsletter] Aug 2023 — Machine Learning Communities: highlights and achievements

[mldp-newsletter]-aug-2023 — machine-learning-communities:-highlights-and-achievements

[MLDP Newsletter] Aug 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!

Photo by James Lee on Unsplash

Keras Community Days are going on!

With the contents from the Keras Sprint, the Keras Community Day campaign by the communities (GDGs, GDSCs, TFUGs and 3p ML communities) is in full swing across the globe!

Keras Community Day by TFUG Surabaya and GDG Surabaya

For instance, TFUG São Paulo hosted Keras Community Day Brazil (video) with Laurence Moroney and Luiz Gustavo. Keras Community Day Surabaya 2023 (photo) was introduced in the local media as a big machine learning event.

Keras Community Day Sub -Saharan Africa (English | French), Keras Community Day Bassam & Day 2 (photo), Keras Community Day Kolkata 2023 (photo), Keras Community Days Hyderabad (photo), Keras Community Day Durg (video), [KCD] Understand the revolution behind AI (video), Keras Community Days Thrissur (photo), Keras Community Day (photo), and others were hosted as impactful events in the each regions.

Keras Community Day by TFUG Jaipur


Introduction to Keras Core: unlocking the power of JAX with Keras Core by ML GDE Aakash Nain (India) and ML GDE Aritra Roy Gosthipaty (India) was an online talk for ML Community members including ML GDEs, TFUG & 3rd party ML community’s organizers. They gave a walkthrough of how to leverage Keras for JAX, how to train neural networks on a single machine and in a distributed setting with multiple GPUs.

Facial, Text Emotion, Speech Emotion Recognition, and MRI Image Classification using Keras and Keras Core by Usha Rengaraju (India) were shared in her Github repository. She also participated in 4 Keras Community Day events as a speakers.

How to build ML Products taking advantage of Gen AI using Keras and Model Garden by ML GDE Nathaly Alarcon Torrico (Bolivia) explained how Gen AI works and showed sample demos with Keras NLP, Keras CV, and Model Garden (Codey).

Transfer Learning With TensorFlow by ML GDE Derrick Mwiti (Kenya) at I/O Extended Nairobi 2023 (hosted by GDG Nairobi) covered transfer learning and what’s new in Cloud and AI/ML.

Keras core: Changes and challenges by ML GDE Jeongkyu Shin (Korea) was a session at I/O Extended Pangyo 2023. He reviewed the recently announced Keras Core, showed how to harness the power of Keras Core with examples, and shared some thoughts on its strengths, weaknesses, and future.

AugMix with KerasCV Breakdown (Part 1): Introduction to AugMix – PyImageSearch

AugMix with KerasCV Breakdown (Part 1): Introduction to AugMix by ML GDE Aritra Roy Gosthipaty (India) and Ritwik Raha (India) is a blog post on AugMix, the augmentation technique, where they break down the paper & its inspiration and discuss the details along with implementation details in KerasCV.

Self Supervised Learning using Keras and Tensorflow by ML GDE Saurav Maheshkar (India) briefly introduced four major families of self supervised learning models along with a walkthrough of a TensorFlow VICReg implementation.

How We See the World: From Equation to Model to Quantum by ML GDE Svetlana Meissner (Germany) at I/O Extended Berlin explained from basic equations to models, linking silicon and human. She also covered how to use Keras and Dialogflow CX to build your own models.

Google I/O Extended Bizerte 2023 by MLAct and GDSC FSB delved into the world of computer vision and the latest updates on TensorFlow, Keras CV, and Keras NLP.


RewardTrainer LoRA — LLM by ML GDE Ertuğrul Demir (Turkey) is a tutorial exploring an alternative approach to solving the Kaggle’s recent Large Language Model Science Exam competition. The method utilizes the increasingly popular reward training style with the help of Low-Rank Adaptation, or LoRA’s.

Kaggle on Twitter: “👀Another beginner-friendly NLP competition: “Contradictory, My Dear Watson” challenges you to detect contradictions & entailments in multilingual text w/ TPUs. Start w/ this KerasNLP tutorial from@kerasteam. (H/T to @asuzsoy for building this comp 👏)https://t.co/S4s4sE8xSB / Twitter”

👀Another beginner-friendly NLP competition: “Contradictory, My Dear Watson” challenges you to detect contradictions & entailments in multilingual text w/ TPUs. Start w/ this KerasNLP tutorial from@kerasteam. (H/T to @asuzsoy for building this comp 👏)https://t.co/S4s4sE8xSB

KerasNLP starter notebook Contradictory DearWatson by ML GDE Alexia Audevart (France) and the Kears team is for a beginner-friendly NLP competition: Contradictory, My Dear Watson. It uses the BERT pretrained model from KerasNLP.

On-device ML

TensorFlow and TensorFlow Decision Forests at an electronic nose by ML GDE George Soloupis (Greece) was a talk about how the TensorFlow ecosystem helps train a model used inside an Android device with data provided from an electronic nose hardware.

Image provided by Pankaj Rai

A session on how to add ML capability to android app by ML GDE Pankaj Rai (India) was an online session organized for university students as a part of India Edu program. He talked about what on-device ML is and its various capabilities like object detection, text recognition, gesture detection, and what frameworks (ML Kit, MediaPipe, TFLite) are available on Android apps along with the code snippets.

Gesture Recognition of 21 gestures using MediaPipe by Kevin Kibe (Kenya) is a project showcasing how to fine-tune a model and perform real-time gesture recognition of 21 different gestures using MediaPipe Model Maker from Google.

I/O Extended 2023 hosted by GDG Mauritius covered on-device machine learning with MediaPipe. I/O Extended Tarija hosted by GDG Tarija also covered how to use MediaPipe + JS to implement machine learning on devices.

LLM / Generative AI


I/O Extended North America was hosted with LLMs and the Future of Programming keynote by Peter Norvig

A Comprehensive Guide about Google Cloud Generative AI Studio by ML GDE Rubens Zimbres (Brazil) includes Python code for servers and clients for many generative AI applications that can be easily tested locally and implemented. The last part presents the fine tuning of a LLM in Generative AI Studio.

PaLM 2: AI in the palm of your hand by ML GDE Jéssica Costa (Brazil) is a lecture given at I/O Extended Pelotas. She talked about how to use PaLM 2 in the Cloud and in general projects. Ethical issues of the evolution of AI were also discussed.

@GoogleDevExpert on Twitter: “Generative AI has been the hot topic at #IOExtended, where ML #GDEs like Kuan are helping devs keep up with Google’s latest tools and innovations in AI!Thank you for sharing your expertise @kuanhoong ❣️ https://t.co/66SVE2SH6z / Twitter”

Generative AI has been the hot topic at #IOExtended, where ML #GDEs like Kuan are helping devs keep up with Google’s latest tools and innovations in AI!Thank you for sharing your expertise @kuanhoong ❣️ https://t.co/66SVE2SH6z

Generative AI with Google PaLM and MakerSuite by ML GDE Kuan Hoong (Malaysia) was a session at I/O Extended Cloud HCM sharing about generative AI with PaLM and MakerSuite.

Learn about the PaLM2 API and LLM by ML GDE Sungmin Han (Korea) was an online session for Korean startup founders and stakeholders. He covered PaLM2 API, VertexAI, large natural language models, and tips for using LLM to help them in real work situations. His other presentation, Sentiment classification of Slack messages with Go and GPT/PaLM2, covered prompt engineering and service development course using PaLM2 and LLMs to create services in Go language.


GenAI — A deep dive by ML GDE Vikram Tiwari (United States) at GDG Seattle I/O Extended 2023 presented the basic building blocks of a generative AI model and how to retain it on your own datasets.

First steps with generative AI on Vertex AI by ML GDE Yannick Serge (Cameroon) at I/O Extended GDG Cloud Sherbrooke, was an online workshop for generative AI. He defined generative AI, explained how it works, and then demonstrated how to use Vertex AI to create and deploy generative AI models.

Bot making Bot: Data to Dialog

Bot making Bot: Data to Dialog by ML GDE Yogesh Kulkarni (India) delves into a transformative workflow that brings this innovation to life, a bot making a bot.

Demystifying Large Language Models and Introducing GCP’s Generative AI offerings by ML GDE Guan Wang (Singapore) at I/O Extended GDG Cloud Da Nang covered the latest development of LLMs including basic ideas behind them, its impact, potential use cases and challenges, and he introduced GCP’s Generative AI offerings including Generative AI Studio, Model Garden and more. His talk included a hands-on session on Generative AI/PaLM API using Cloud Boost Skills credits.

Have you used BARD today? by ML GDE Fillipe Dornelas (Brazil) talked about what he saw at Google I/O Connect Miami and what’s new in ML. He also shared how to use each ML tool presented at the event.

Generative AI on GCP by ML GDE Yüksel Tolun (Turkey) at I/O Extended Ankara introduced Gen AI features on GCP, started from the basic concepts of Gen AI App building to Gen App AI Builder, Vertex AI w/ Gen AI support, some PaLM APIs, and langchain magic. He also gave the same talk in I/O Extended Izmir and Gaziantep.


Google’s Duet AI vs ChatGPT Enterprise, Big Funding Rounds, GM & More by ML GDE Allen Firstenberg (United States) is a conversation video about the state of generative AI in general and the latest GenAI news from Google Cloud Next. He also presented Using LLMs to Bridge the Fuzzy Human / Digital Computer Boundary at I/O Extended North America, showing how LLMs can be used to help computers “understand” a fuzzy human request into a discrete data structure that computers can process. He showed examples using LangChain JS to turn a user’s request into a SQL query, and then into human-digestible results.

Fine-tuning LLMs with TensorFlow and GCP by ML GDE Karthik Muthuswamy (Germany) shared how to use TensorFlow and GCP to fine-tune LLMs for various applications.

India Edu program’s Tech Talks for Educators series with 1.3K educators registrations. Several educators hosted workshops for students after the session.

ML Research

Two paper reviews by ML GDE Grigory Sapunov (UK) for HyperDreamBooth: HyperNetworks for Fast Personalization of Text-to-Image Models and Resurrecting Recurrent Neural Networks for Long Sequences shared on his social channel.

Training RecSys retrieval models with TensorFlow and Merlin — EvalRS workshop at KDD’23 by ML GDE Gabriel Moreira (United States) was about how to build RecSys models with TensorFlow and Merlin libraries presented to the hackathon competitors of EvalRS workshop at KDD.

Modeling The Brain’s Response To Natural Scenes by Henry Ndubuaku (UK) is a research on training artificial neural networks to map visual stimuli to the corresponding brain responses which often requires working with multiple subjects, high-dimensional data, and thus expensive computation. He proposed improving efficiency with dimensionality reduction and general pre-training on subjects, as well as novel racing loss and ground truth injection functions. He used JAX/Flax for the research.


So what is Dynamic Token Routing? by ML GDE Ritwik Raha (India) is a video summary of the paper, DiT: Efficient Vision Transformers with Dynamic Token Routing, making it easier for anyone to read and understand the concepts and math in a jargon-free way.

Activities by ML Frameworks


#BigQuery # Duet AI #Vertex AI GDG Cloud Next Extended Singapore hosted by GDG Cloud Singapore covered AI Lakehouse, LLMs in BigQuery, boosting productivity with GenAI, Duet AI and Vertex AI.

#Visual Blocks Visual Blocks: Democratizing AI in the Industry by ML GDE Hugo Zanini (Brazil) showed how Visual Blocks can be used as a tool for generating solutions in the industry. He shared how to create ML models from scratch in the tool and how to use pre-built nodes. The main tasks covered were related to CV and LLMs.

#TensorFlow Tweet Image Maker: Building a multimodal recommender system to recommend engaging images/hashtags for Tweets by ML GDE Esther Irawati Setiawan (Indonesia) at I/O Extended Semarang introduced Social Network Analysis (SNA) in collaboration with ML and TensorFlow to participants. She elaborated on how to build a multimodal recommender system with ML approach using CLIP and TensorFlow Recommender.

#TensorFlow #Go TensorFlow for Go, Let’s make an AI service in Go by ML GDE Sungmin Han (Korea) was for beginners to create AI services. He covered how to run the TensorFlow framework in the Go language for learning and serving.

Trabalhando com Tensorflow Decision Forests-Parte 2

#TensorFlow Working with Tensorflow Decision Forests-Part 2 by ML GDE Jéssica Costa (Brazil) is a series of posts giving a vision about unbalanced data and results interpretation in TensorFlow Decision Forests.

#TensorFlow.js AI Fundamentals Bootcamp: Introduction to Machine Learning by Patty O’Callaghan at GDG Barcelona & GDG Glasgow was a series of events introducing artificial neural networks and TensorFlow.js

#SimpleML I/O Extended 2023 by GDG Cloud Kolkata introduced Simple ML that helps using machine learning without coding or prior ML expertise.

#BigQuery I/O Extended — Cloud Edition hosted by GDG Cloud Santiago de Chile & GDG Santiago de Chile covered the latest news in Google technology including Scaling large language models in the Cloud and Data modeling and prediction with BigQuery ML.

#DialogFlow FitBit Chatbot Dialogflow by ML GDE Rubens Zimbres (Brazil) is a video introducing a chatbot made with Dialogflow CX. This chatbot uses generative AI to provide answers regarding the FitBit device’s manual.


Margaret M on Twitter: “My story “Being an #ML #GDE” with my cool projects in this amazing community! #TensorFlow #TFHub #Keras #MediaPipe #TFLite #MLKit #GenAI #GoogleCloud Thanks @ksoonson, @heeya_ML, Nari, @GoogleDevExpert, #MLGDEs & Google product teams for collaboration!https://t.co/uRKZaDhJAz pic.twitter.com/sXSjaQVN3m / Twitter”

My story “Being an #ML #GDE” with my cool projects in this amazing community! #TensorFlow #TFHub #Keras #MediaPipe #TFLite #MLKit #GenAI #GoogleCloud Thanks @ksoonson, @heeya_ML, Nari, @GoogleDevExpert, #MLGDEs & Google product teams for collaboration!https://t.co/uRKZaDhJAz pic.twitter.com/sXSjaQVN3m

Being an ML Google Developer Expert by ML GDE Margaret Maynard-Reid (United States) is her experience of being an ML GDE and cool projects she’s worked on as part of the amazing ML community.

Collaborate, Innovate, Elevate: The Power of Communities by Aashi Dutt at I/O Extended Jalandhar was a talk about how to collaborate within communities, innovate and elevate each other.


Introducing how Google Technology is Improving the Machine Learning Ecosystem by ML GDE Esther Irawati Setiawan(Indonesia) and Googler Janise Tan was an interview by a radio station having a fan base of more than 1 million listeners every month.

[MLDP Newsletter] Aug 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.

Leave a Reply

Your email address will not be published. Required fields are marked *

Previous Post

19 SEO Tips Straight From the Mouths of HubSpot’s SEO Team

Next Post

What Is an NFT Marketplace and How to Create an NFT Marketplace?

Related Posts