Episode 68 of The Data Science Podcast with Fexingo dives into how data scientists are using embeddings — dense vector representations of text, images, and other data — to power semantic search and information retrieval. Lucas and Luna explore a concrete case: a mid-sized e-commerce company that replaced its keyword-based search with embeddings and saw a 40% improvement in product discovery. They break down what embeddings are, how models like BERT generate them, and why cosine similarity is the go-to metric for retrieval. The hosts also discuss trade-offs like the cost of generating embeddings at scale and the rise of hybrid search systems that combine keywords with embeddings to get the best of both worlds. If you've ever wondered how modern search engines understand intent rather than just matching words, this episode gives you a clear, example-driven explanation.

#DataScience #Embeddings #SemanticSearch #MachineLearning #NLP #VectorSearch #CosineSimilarity #BERT #InformationRetrieval #Ecommerce #RAG #Technology #Podcast #FexingoBusiness #BusinessPodcast #Search #DeepLearning #DataScientists

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna dive into how graph neural networks (GNNs) are transforming fraud detection in financial systems. They explore a real case from a major European bank that deployed GNNs to catch synthetic identity fraud — a scheme that cost U.S. lenders an estimated $6 billion in 2025. Lucas breaks down why traditional machine learning models fail on relational fraud patterns, how GNNs exploit transaction graphs, and the surprising finding that adding just two hops of neighbor information improved recall by 40%. Luna asks the tough questions about computational cost and explainability, and they discuss practical tools like PyTorch Geometric and DGL. If you're a data scientist looking for a cutting-edge application of deep learning on graphs, this episode is for you.

#GraphNeuralNetworks #FraudDetection #DataScience #MachineLearning #Technology #Business #Finance #SyntheticIdentityFraud #PyTorchGeometric #DGL #AnomalyDetection #FinancialServices #DeepLearning #GraphAnalytics #FexingoBusiness #BusinessPodcast #TheDataSciencePodcast #LucasAndLuna

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna dive into counterfactual explanations—a technique that helps data scientists make machine learning models more interpretable by showing what would need to change for a prediction to flip. They walk through a concrete example from a credit approval model at a European fintech, where a loan denial was explained by showing that increasing income by $5,000 would have changed the outcome. The hosts discuss how counterfactuals differ from traditional feature importance, the trade-offs between proximity and plausibility, and the emerging regulatory push under the EU AI Act that makes this approach increasingly relevant. They also touch on open-source tools like What-If Tool and ALIBI that practitioners can use today. No fluff—just a focused look at one practical technique for building trust in black-box models.

#CounterfactualExplanations #ExplainableAI #XAI #ModelInterpretability #MachineLearning #DataScience #AIEthics #CreditApproval #EUAIAct #WhatIfTool #ALIBI #LIME #SHAP #Technology #Podcast #FexingoBusiness #BusinessPodcast #DataDriven

Keep every episode free: buymeacoffee.com/fexingo

In this episode of The Data Science Podcast, Lucas and Luna explore survival analysis—a statistical method originally developed for medical research—and how data scientists are applying it to predict customer churn and retention. They walk through a concrete example: a subscription service using Kaplan-Meier curves and Cox proportional hazards models to understand when and why customers cancel. Lucas explains the concept of the 'hazard function' and how it differs from traditional churn models, while Luna challenges him on interpretability. The hosts also touch on real-world applications in SaaS, insurance, and telecom, and discuss the difference between censored and uncensored data. By the end, you'll understand why survival analysis offers a more nuanced view of retention than simple binary classification.

#SurvivalAnalysis #CustomerRetention #ChurnPrediction #KaplanMeier #CoxProportionalHazards #DataScience #MachineLearning #Statistics #SaaS #SubscriptionBusiness #CensoredData #HazardFunction #BusinessAnalytics #PredictiveModeling #Technology #FexingoBusiness #BusinessPodcast #DataDriven

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna dive into the practical side of recommendation systems, focusing on how data scientists at companies like Spotify and Netflix move beyond simple collaborative filtering. They explore the evolution from matrix factorization to two-tower neural networks, the critical role of negative sampling, and why offline metrics often don't match real-world success. Specific numbers include a 2017 Netflix prize insight and a 20 percent improvement in user engagement from Spotify's deep learning model. The conversation also touches on the cold start problem and how to handle new users or items. A must-listen for anyone building or evaluating recommender systems in production.

#RecommendationSystems #MachineLearning #CollaborativeFiltering #MatrixFactorization #TwoTowerNetworks #NegativeSampling #NetflixPrize #Spotify #ColdStart #DataScience #Technology #FexingoBusiness #BusinessPodcast #DataDriven #Personalization #DeepLearning #OfflineMetrics #UserEngagement

Keep every episode free: buymeacoffee.com/fexingo

In this episode of The Data Science Podcast, Lucas and Luna dive into differential privacy—a mathematical framework that lets data scientists extract useful insights from datasets without revealing information about any single individual. They walk through a concrete example from the 2020 US Census, where the Census Bureau added calibrated noise to protect respondent privacy while preserving statistical accuracy. Lucas explains the epsilon parameter, the trade-off between privacy and utility, and how companies like Apple and Google have deployed differential privacy in products like iOS keyboard predictions and Chrome telemetry. Luna asks whether the noise ruins the data for small subgroups and whether this is really a silver bullet for privacy. The episode closes with practical guidance on implementing differential privacy in production using open-source libraries like Google's Differential Privacy library and the OpenDP platform. If you work with sensitive data and want to learn how to guard against re-identification attacks without sacrificing analytical value, this episode is for you.

#DifferentialPrivacy #DataPrivacy #CensusBureau #Apple #Google #EpsilonParameter #StatisticalNoise #PrivacyUtilityTradeoff #OpenDP #ReIdentificationAttack #MachineLearning #DataScience #Technology #BusinessPodcast #FexingoBusiness #TheDataSciencePodcast #TechPodcast #PrivacyPreservingML

Keep every episode free: buymeacoffee.com/fexingo

Episode 62 of The Data Science Podcast dives into the operational side of machine learning: MLOps. Lucas and Luna explore why so many models never make it to production, and how tools like feature stores, model registries, and automated pipelines keep deployed models accurate and reliable. They walk through a real case from a mid-sized fintech company that cut model deployment time from weeks to hours using CI/CD for ML. Along the way, they touch on monitoring drift, versioning data, and the cultural shift from research to engineering. If you've ever wondered why building a model is only half the battle, this episode gives you the other half.

#DataScience #MLOps #MachineLearning #AIEngineering #ModelDeployment #FeatureStore #ModelRegistry #CICD #DataPipeline #ModelDrift #ProductionML #Fintech #Tech #BusinessPodcast #FexingoBusiness #Podcast #DataEngineering #DevOps

Keep every episode free: buymeacoffee.com/fexingo

Retrieval-augmented generation, or RAG, is reshaping how companies deploy large language models without retraining. In this episode, Lucas and Luna drill into the data-science architecture behind RAG: how vector databases encode semantic meaning, why cosine similarity beats keyword search, and what a production RAG pipeline looks like at a mid-size fintech startup. They walk through a concrete example—building a customer-support bot for a payments company—showing where embedding models, chunking strategies, and approximate nearest-neighbor search come into play. Lucas breaks down the trade-off between accuracy and latency, and Luna questions whether RAG is just a band-aid for models that can't reason. Tune in for a grounded look at the database layer that's quietly powering the next wave of AI applications.

#VectorDatabases #RAG #RetrievalAugmentedGeneration #LLM #Embeddings #CosineSimilarity #ApproximateNearestNeighbor #DataScience #MachineLearning #NLP #AIArchitecture #CustomerSupport #Fintech #Pinecone #Weaviate #Milvus #FexingoBusiness #TechnologyPodcast

Keep every episode free: buymeacoffee.com/fexingo

Anomaly detection is one of the most practical applications of data science in industry today. In episode 60 of The Data Science Podcast, Lucas and Luna dive into how modern data scientists build real-time anomaly detection systems that catch fraud, equipment failures, and network intrusions before they escalate. They walk through a concrete example: a mid-sized e-commerce company that reduced chargeback fraud by 37% using an isolation forest model on streaming transaction data. They discuss trade-offs between sensitivity and false positives, the role of feature engineering in time-series data, and why 'weird' isn't always bad. Along the way, they touch on how tools like Apache Kafka and streaming analytics platforms have made real-time detection accessible beyond big tech. If you've ever wondered how data scientists spot the needle in the haystack — without drowning in alerts — this episode will give you both the mental model and the practical gotchas.

#AnomalyDetection #RealTimeAnalytics #IsolationForest #FraudDetection #DataScience #MachineLearning #StreamingData #ApacheKafka #FeatureEngineering #TimeSeries #FalsePositives #Ecommerce #Technology #DataDriven #FexingoBusiness #BusinessPodcast #DataSciencePodcast #ModelDeployment

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna explore why traditional A/B testing falls short when data is sparse or you need to compare more than two variants. They walk through a concrete example: a subscription site testing four pricing tiers with only 2,000 visitors per variant. Using Bayesian methods with a Beta-Binomial conjugate prior, the team got actionable probability estimates in three days instead of waiting two weeks for frequentist significance. They cover prior selection, posterior distributions, and the practical business trade-off between speed and certainty. Tune in to learn why Bayesian A/B testing is becoming standard at companies like Booking.com and Etsy.

#BayesianABTesting #DataScience #A_BTesting #BetaBinomial #PriorDistribution #PosteriorProbability #FrequentistVsBayesian #BusinessDecisions #Experimentation #PricingOptimization #Bookingcom #Etsy #Technology #MachineLearning #Statistics #FexingoBusiness #BusinessPodcast #DataDriven

Keep every episode free: buymeacoffee.com/fexingo