The Data Science Podcast with Fexingo: Analytics, Machine Learning, and Data-Driven Conversations

In this episode, Lucas and Luna dive into conformal prediction, a model-agnostic framework that gives machine learning models reliable uncertainty estimates without sacrificing coverage guarantees. They discuss how it works — using a calibration set to produce prediction sets with a user-specified confidence level — and walk through a concrete example from medical imaging where a model flags skin...

In this episode, Lucas and Luna explore how knowledge distillation allows data scientists to compress large neural networks into smaller, faster models without catastrophic accuracy loss. They break down the teacher-student training paradigm, using real examples from Google's DistilBERT — which shrank BERT by 40% while retaining 97% of its language understanding — and NVIDIA's work compressing vis...

Episode 76 of The Data Science Podcast dives into causal forests, a powerful tree-based method for estimating heterogeneous treatment effects. Lucas and Luna unpack how this technique helps data scientists answer not just 'does a treatment work?' but 'who benefits most?' using real-world examples from personalized medicine and targeted marketing. They walk through the intuition behind honest split...

Episode 75 of The Data Science Podcast dives into Temporal Fusion Transformers (TFT), a deep learning architecture that's changing how data scientists handle multi-horizon time series forecasting. Lucas and Luna break down how TFT combines interpretable components like variable selection networks with attention mechanisms to produce accurate, explainable forecasts. They walk through a real-world c...

Lucas kicks off Episode 74 of The Data Science Podcast with a specific problem: a model is retrained every night, but each team rebuilds the same features from scratch, wasting compute and introducing inconsistency. He introduces feature stores — centralized repositories for defining, sharing, and serving features — as the solution. Luna asks how feature stores handle time-travel for point-in-time...

Lucas and Luna explore how data scientists apply counterfactual regret minimization (CFR) to solve strategic decision-making in games like poker and beyond. They break down the concept using the concrete example of Pluribus, the AI that beat world-class poker players in no-limit Texas Hold'em. Lucas explains how CFR iteratively evaluates decisions by comparing actual outcomes to 'what if' scenario...

Data augmentation is a cornerstone of modern machine learning, but traditional methods like rotation, cropping, and synonym replacement have limits. In this episode, Lucas and Luna explore how large language models are changing the game. They discuss a real case from a marketing analytics startup that used GPT-4 to generate synthetic customer reviews for training a sentiment classifier. Lucas expl...

Episode 71 of The Data Science Podcast explores active learning — a technique where models strategically query the most informative data points for human labeling, drastically reducing annotation costs. Lucas and Luna walk through a real-world example from medical imaging: training a diagnostic model to detect lung nodules with 80 percent less labeled data than traditional approaches. They explain...

In episode 70 of The Data Science Podcast, Lucas and Luna explore Gaussian processes: a powerful Bayesian method for quantifying uncertainty in predictions. They anchor the discussion on a concrete use case: predicting manufacturing yield for a semiconductor fabrication plant, where knowing the confidence interval matters as much as the point estimate. Lucas explains how Gaussian processes differ...

Episode 69 of The Data Science Podcast with Fexingo dives into contrastive learning, a self-supervised technique reshaping computer vision. Lucas and Luna break down how SimCLR and MoCo let models learn visual representations without labeled data, using the example of a medical imaging startup that cut annotation costs by 80%. They explore the core idea of pulling similar images together and pushi...

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...

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 G...

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 ch...

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. Lu...

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 numb...

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 stati...

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 h...

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 exa...

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 cha...

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 wee...

Episode 58 of The Data Science Podcast explores how data scientists are turning to synthetic data when real-world data is scarce, expensive, or privacy-sensitive. Lucas and Luna break down a concrete case: how a European insurance company used the Synthpop library in R to generate synthetic claims data, training a fraud detection model that outperformed their original one. They discuss the core te...

Episode 57 of The Data Science Podcast dives into double machine learning (DML), a technique that estimates causal effects from observational data without assuming a linear model. Lucas and Luna walk through a real-world example: a health-tech startup using DML to determine whether a new wellness program reduces employee churn. They explain why traditional regression falls short, how DML combines...

When a new product launches with zero user history, recommendation systems and personalization engines face the 'cold start' problem — they have no data to learn from. In this episode, Lucas and Luna explore how data scientists are using transfer learning to jump-start predictions without waiting for users to generate behavior. They walk through a real example from an e-commerce startup that used...

Episode 55 of The Data Science Podcast with Fexingo explores knowledge graphs — a technique that links siloed data sources into a unified, queryable structure. Lucas and Luna walk through the basics of how knowledge graphs represent entities and relationships, then dive into a concrete example: how a hospital chain used a knowledge graph to combine electronic health records, lab results, and socia...

In episode 54 of The Data Science Podcast, Lucas and Luna explore how companies are moving beyond correlation to use causal inference for real business impact. They break down a concrete case: how a major retailer used a natural experiment — a regional shipping delay caused by a freak snowstorm — to estimate the true revenue effect of faster delivery promises. Lucas explains the difference between...