17º encontro da comunidade Data Science Joinville

https://www.meetup.com/pt-BR/data-science-joinville/events/262841253/

Transcript

1. View Slide

2. View Slide

3. THE MOST DIVERSE CONFERENCE ON ML
   Dan Nichol
   Program Chair,
   AnalyticsFC
   Ana Paula Appel
   Researcher and Data
   Scientist, IBM Research
   Louis Dorard
   General Chair, PAPIs
   

   View Slide

4. Julien Simon
   AI Evangelist, AWS
   https://aws.amazon.com/deepracer/
   Reinforcement Learning for devs: toolkits, and application to autonomous race cars
   ● SageMaker <3
   

   View Slide

5. Leela Senthil Nathan
   Software Engineer, Stripe
   ● Apresentação
   ● Fraude de Cartão de Crédito
   ● Separação de Samples de Falso Positivos X Falsos
   Negativos
   ● O problema do desbalanceamento de amostras
   Lessons Learned from Building a Credit Card Fraud Model
   

   View Slide

6. Bias and bugs: implementing recommendations
   Guilherme Silveira
   Head of Education, Alura
   ● Apresentação
   ● Recomendação de cursos;
   ● Vies e outliers;
   ● Correlação x Casualidade;
   ● A/B bem feitos
   

   View Slide

7. TensorFlow image inferencing: an adventure in Python and Go
   ● Análise de imagens de pulmão
   ● Tensorflow API: Python ou Go?
   ● Problemas com GO
   ● Python + Keras
   ● Integração GO (HTTP, Flask and RabbitMQ)
   Vitor De Mario
   Tech Lead, NeuralMed
   

   View Slide

8. ● Test A/B;
   ● Relevância Estatistica;
   ● Comparação de Forescast e A/B
   ● Sazonalidade;
   ● Prophet
   Where are the gains: Should I use A/B Tests for forecasting?
   Eder Martins
   Data scientist, SEEK
   

   View Slide

9. ● Apresentação
   ● DS e ML não precisa ser
   complicado;
   ● Soluções Simples
   ● Tecnico x Comum
   ● Comunicação
   Adauto Braz
   Data Scientist, Stoodi
   I know what you did last session: clustering users with ML
   

   View Slide

10. Using Machine Learning to recommend jobs in User Cold Start
    Andryw Marques
    Data Scientist, SEEK
    ● Apresentação;
    ● Vaga x Candidato;
    ● Features simples;
    ● Teste A/B + Feature Flag;
    ● Pesos para Features x Vaga
    

    View Slide

11. A Clinical Application of Deep Learning for NLP with Word-Embeddings
    Arnon Santos
    Data Scientist,
    Junto Seguros
    ● Apresentação;
    ● Catálogo de Protuários;
    

    View Slide

12. Time Series Forecasting for Cloud Resources Provisioning
    Leonardo Neri
    AI Manager, Accenture
    ● Provisionamento de máquinas;
    ● Forescast por Log;
    ● Prophet;
    ● “Eventos” fake simular
    sazonalidade;
    

    View Slide

13. How Big Data Powers Ambev’s Sales Machine - sponsored by Big Data
    Gustavo Ioschpe
    Founder & CEO, Big Data
    ● Original onde só tinha Brahma;
    ● Logistica;
    ● Foto Local x Tipo de Bebida
    

    View Slide

14. How Machine Learning is Transforming the Way iFood Runs its Logistic - sponsored by Movile
    Arnaud Seydoux
    Logistic IA manager, iFood
    ● Logistica de Entregas;
    ● Tempo de Preparo x Tempo de Entrega;
    ● Predição Motoboy;
    ● Predição Cozinha;
    ● Escalar x Modelo atualizado
    

    View Slide

15. Panel: Diversity in ML
    Larissa Lautert
    Data Scientist, Linx Impulse
    Larissa Lautert
    Data Scientist, Linx Impulse
    Ana Paula Appel
    Researcher and Data Scientist, IBM
    Research
    Ricardo Herrmann
    Artificial Intelligence Engineer, Olivia
    AI
    

    View Slide

16. Ludwig, a Code-Free Deep Learning Toolbox
    Piero Molino
    Sr. Research Scientist, Uber AI
    ● https://uber.github.io/ludwig/
    

    View Slide

17. DataOps architecture for Machine
    Learning - brought to you by everis
    Apresentação
    Carlos Porto Filho
    Data Scientist, Everis
    Gustavo Castilhos
    Data Architect, everis
    

    View Slide

18. AI culture and semi-autonomous review approval - sponsored by Dafiti
    Ricardo Savii
    Data Scientist, Dafiti Group
    Apresentação
    

    View Slide

19. BERT: Multi-language approach for Q&A and NLP Applications
    ● Bidirectional Encoder Representation of Transformers (BERT)
    ○ Modelo de representação de linguagem
    ○ Modelo pré treinado pode receber um ajuste-fino com
    apenas uma camada de saída adicional
    ○ “While the empirical results are strong, in some cases
    surpassing human performance ...”
    ● Question & Answers e outros problemas de NLP
    Horst Rosa Erdmann
    Lead Data Scientist, Everis
    

    View Slide

20. Fklearn: A functional library for machine learning
    Henrique Lopes
    Machine Learning Engineer, Nubank
    ● https://github.com/nubank/fklearn/blob/master/docs/source/exam
    ples/fklearn_overview.ipynb
    

    View Slide

21. Multitask convolutional neural networks: saving GPU time
    ● Companhia de varejo com mais 250 milhões de produtos
    ● CNN (Convolutional Neural Networks)
    ● Single task classifier (mais classificadores)
    ● Multi task classifier (All in one go)
    Paulo Eduardo Sampaio
    Data science specialist, McKinsey &
    Company
    

    View Slide

22. Paulo Eduardo Sampaio
    Data science specialist, McKinsey &
    Company
    Multitask convolutional neural networks: saving GPU time
    ● Companhia de varejo com mais 250 milhões de produtos
    ● CNN (Convolutional Neural Networks)
    ● Single task classifier (2 classificadores)
    ● Multi task classifier (All in one go)
    

    View Slide

23. Reproducibility with Data Version Control
    Victor Villas Bôas Chaves
    Data Engineer, Gupy
    ● Apresentação
    ● https://dvc.org/
    

    View Slide

24. Validating models in the real world
    Luis Moneda
    Data Scientist,
    Nubank
    Apresentação
    

    View Slide

25. ETL Orchestration with AWS Glue and AWS Step-functions
    Alexsandro Francisco dos
    Santos
    Data Engineer, Stoodi
    

    View Slide

26. How Grupo ZAP is using data to
    empower real estate buyers,
    sellers, and renters in Brazil -
    sponsored by Grupo ZAP
    Lucas Vargas
    CEO, Grupo ZAP
    

    View Slide

27. Machine Learning for Natural Resources
    Bianca Zadrozny
    Research Manager, IBM
    ● Mineração de Ouro;
    ● Geologista x ML;
    ● Geologista e ML;
    ● Grande área x Pepita;
    ● 3D x 2D
    

    View Slide

28. Panel: How much should we
    care about model
    interpretability?
    Renato Vicente
    Chief Scientist, Serasa
    Experian
    Sandor Caetano
    Chief Data Scientist,
    IFood / Movile
    Ivan Marin
    Data Scientist/System
    Architect, Daitan Group
    

    View Slide

29. ● Youtube
    ● Twitter
    ● Site
    ● Medium
    ● Resumo 1
    ● Resumo 2
    ● #papisio
    ● https://www.linkedin.com/in/gabubellon/
    ● https://www.linkedin.com/in/fachini/
    ● https://www.linkedin.com/in/gplichoski/
    

    View Slide