• Design and maintain data pipelines using Airflow (DAGs) to support fraud detection workflows • Process and analyze large volumes of transactional data using Spark • Develop, evaluate, and monitor fraud risk scoring models • Perform exploratory analyses to identify fraud patterns, anomalies, and emerging risk behaviors • Create and maintain dashboards to monitor fraud KPIs (approval rate, chargeback rate, fraud rate, false positive rate, among others) • Collaborate with Product, Payments, and Risk teams to evolve fraud prevention strategies • Contribute to MLOps and ML Engineering practices for deployment and monitoring of fraud models

• Design and own analytical frameworks that guide the evolution of our auction and bidding systems, including CPC pricing, bid landscape modeling, and reserve price optimization. • Run rigorous experiments to evaluate the impact of auction mechanism changes—such as bid shading, second-price vs. generalized second-price auctions, and dynamic pricing—and translate results into actionable recommendations. • Develop statistical and machine learning models to forecast bid distributions, predict click-through rates, and optimize auction efficiency across advertiser and consumer dimensions. • Build simulations to model auction dynamics and stress-test proposed mechanism changes before live deployment. • Partner closely with Engineering, ML, and Product to translate business and advertiser needs into well-specified model and system requirements. • Enable objective decision-making by building dashboards and monitoring frameworks that surface auction health, advertiser performance, and CPC trends to stakeholders across the organization. • Apply expertise in causal inference, mechanism design, and econometrics to ensure our auction systems are fair, efficient, and aligned with marketplace goals. • Present analytical findings and strategic recommendations in a compelling way to influence Instacart's leadership and Ads go-to-market teams.

• Design and own analytical frameworks that guide the evolution of our auction and bidding systems, including CPC pricing, bid landscape modeling, and reserve price optimization. • Run rigorous experiments to evaluate the impact of auction mechanism changes—such as bid shading, second-price vs. generalized second-price auctions, and dynamic pricing—and translate results into actionable recommendations. • Develop statistical and machine learning models to forecast bid distributions, predict click-through rates, and optimize auction efficiency across advertiser and consumer dimensions. • Build simulations to model auction dynamics and stress-test proposed mechanism changes before live deployment. • Partner closely with Engineering, ML, and Product to translate business and advertiser needs into well-specified model and system requirements. • Enable objective decision-making by building dashboards and monitoring frameworks that surface auction health, advertiser performance, and CPC trends to stakeholders across the organization. • Apply expertise in causal inference, mechanism design, and econometrics to ensure our auction systems are fair, efficient, and aligned with marketplace goals. • Present analytical findings and strategic recommendations in a compelling way to influence Instacart's leadership and Ads go-to-market teams.

• Desenvolver modelos de classificação e regressão para previsão de: probabilidade de êxito, tipo de sentença, comportamento do processo ao longo da tramitação; • Criar e testar diferentes abordagens: modelos clássicos (Logistic Regression, XGBoost, Random Forest), modelos baseados em NLP, embeddings, Transformers e modelos híbridos (estatística + semântica textual); • Analisar petições iniciais e outras peças para extrair: entidades jurídicas (NER), emas, fundamentos, pedidos e características relevantes, similaridades entre processos; • Criar features semânticas e estruturadas para alimentar os modelos; • Atualizar-se em metodologias e algoritmos de vanguarda para otimizar acurácia e demais métricas relevantes, no contexto de predição de sentenças jurídicas; • Realizar análises exploratórias profundas (EDA) de dados jurídicos fornecidos pelo sistema; • Testar diferentes estruturas de dados para melhorar performance dos modelos; • Construção de provas de conceito (POCs) com LLMs voltadas para: sumarização de peças, andamentos e decisões; geração assistida de textos jurídicos (minutas, pareceres preliminares etc.); apoio aos procuradores com agentes especializados; • Trabalhar com RAG (Retrieval-Augmented Generation) conectado a bases jurídicas; semantic Search para captura de entendimento jurídico; ajuste fino (fine-tuning) ou instrução de modelos generativos para tarefas específicas e benchmarking de modelos abertos e proprietários; • Validar modelos usando métricas robustas (AUC, F1, KS etc.); • Aplicar técnicas de explicabilidade/representatividade das variáveis decisoras; • Gerar relatórios técnicos explicando: metodologia, interpretação dos resultados, relevância das variáveis, limitações e próximos passos, documentar versões, experimentos e procedimentos de treinamento; • Especificar requisitos técnicos para integração dos modelos; • Entregar artefatos e orientações técnicas sobre input/output dos modelos;