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

Episode 53 of The Data Science Podcast dives into federated learning — a technique that trains machine learning models across decentralized data without ever moving sensitive information to a central server. Lucas and Luna break down how Google used it to improve keyboard predictions on Gboard without seeing what you type, and how Apple applies it to Siri without uploading your voice. They discuss...

In this episode of The Data Science Podcast, Lucas and Luna explore how reinforcement learning (RL) is transforming dynamic pricing strategies. Using the example of a major ride-hailing company, they break down how RL algorithms learn to set prices in real time by balancing exploration (testing new price points) and exploitation (using known optimal prices). Lucas explains the core RL concepts of...

Churn prediction is one of the most common — and most frustrating — problems data scientists face. In this episode, Lucas and Luna dig into why many churn models fail in production and what separates the ones that actually reduce customer loss. They walk through a concrete example from a mid-size telecom company that cut churn by 14 percent in six months by focusing on the right features and the r...

In this milestone 50th episode, Lucas and Luna explore active learning — a machine learning paradigm where the model itself chooses which data points to label, dramatically reducing manual annotation costs. They break down the core idea using a concrete example: training a fraud detection model for a payment processor processing 10 million transactions per day. Lucas explains uncertainty sampling,...

When your dataset outgrows a single machine, what do you do? In this episode, Lucas and Luna explore how data scientists use distributed computing frameworks like Apache Spark and Dask to process terabytes of data without crashing their laptops. They break down the key concept of data partitioning, explain why MapReduce is still relevant, and walk through a real example of how a mid-sized e-commer...

This episode of The Data Science Podcast dives into TinyML—the practice of running machine learning models on microcontrollers and edge devices with limited power and memory. Lucas and Luna explore a specific case: a smart agriculture startup that used a TensorFlow Lite model on an Arduino board to detect crop disease in real time, processing images locally without sending data to the cloud. They...

In this episode, Lucas and Luna dive into the growing role of natural language processing in detecting online misinformation. They explore a specific case study: how DataGPT, a startup, built a fact-checking bot that flags false claims in real-time on social media. Lucas breaks down the technical stack — transformer models like BERT fine-tuned on fact-checking datasets, with a focus on stance dete...

Episode 46 of The Data Science Podcast with Fexingo explores how graph neural networks (GNNs) are transforming fraud detection in financial services. Lucas and Luna break down a real case: how a major European bank used GNNs to catch a money-laundering ring that rule-based systems missed. They discuss why traditional fraud models fail with relational data, how GNNs represent transactions as a grap...

Episode 45 of The Data Science Podcast explores the emerging practice of counterfactual explanations — 'what-if' scenarios that help end users understand why a machine learning model made a particular decision. Lucas and Luna walk through a concrete example from the lending industry: a small business owner named Maria whose loan application was denied by an automated risk model. Instead of a black...

Episode 44 of The Data Science Podcast with Fexingo dives into Shapley Values, a game-theory approach to explaining individual model predictions. Lucas and Luna break down how this method works, why it's more principled than feature importance rankings, and where it falls short in practice. They walk through a concrete example from credit scoring, explaining how a single applicant's denial can be...

Episode 43 of The Data Science Podcast with Fexingo. Lucas and Luna explore how machine learning is transforming medical diagnosis — specifically, the case of a deep learning model developed at Stanford that detects skin cancer with accuracy comparable to board-certified dermatologists. Published in Nature in 2017, the model was trained on nearly 130,000 images representing over 2,000 diseases. Lu...

Lucas and Luna walk through the real-world process of building a recommendation engine, using the open-source MovieLens dataset as their running case. They cover collaborative filtering, matrix factorization, the cold-start problem, and the engineering trade-offs between offline accuracy and online performance. Lucas explains why the Netflix Prize algorithm never made it into production, and Luna...

Lucas and Luna tackle the often-overlooked chasm between a promising prototype and a production model that actually drives business value. They dive into a 2023 Gartner finding that 85 percent of data science projects never reach deployment, and unpack the three biggest culprits: misaligned incentives between data scientists and engineers, brittle code that can't scale, and missing infrastructure...

Lucas and Luna dive into synthetic control methods, a causal inference technique that data scientists use when A/B testing is impossible. They walk through a concrete example: how a mid-sized retailer used synthetic controls to measure the revenue impact of opening a new physical store, using a weighted combination of similar stores as a counterfactual. Lucas explains the math behind the method—ma...

Episode 39 of The Data Science Podcast explores the growing discipline of model monitoring in production. Lucas and Luna discuss why many data science teams still treat monitoring as an afterthought, how silent failures erode business trust, and what tools like Evidently AI, WhyLabs, and custom dashboards are doing about it. They walk through a real example from a fintech lending platform where a...

Lucas and Luna dive into Bayesian A/B testing, a method that's quietly replacing traditional frequentist approaches in data science. They break down how it works, why it's more intuitive, and where it falls short. The episode centers on a real case: how a major retailer used Bayesian testing to optimize their checkout flow, cutting decision time from weeks to days. Lucas explains the math behind p...

In this episode of The Data Science Podcast, Lucas and Luna dive into Spotify's recommendation engine — not the playlist curation you already know, but the predictive models that identify tracks you haven't heard yet. They break down the specific machine learning techniques behind Spotify's 'Discover Weekly' and 'Release Radar,' focusing on collaborative filtering, natural language processing of a...

Netflix tests millions of thumbnail combinations daily using multi-armed bandit algorithms — a technique smarter than traditional A/B testing. In this episode, Lucas and Luna break down how contextual bandits balance exploration and exploitation to serve the most clickable image for each user, without wasting traffic on losing variants. They walk through the epsilon-greedy approach, why Netflix ab...

Episode 35 of The Data Science Podcast tackles the messy reality of algorithmic fairness. Lucas and Luna explore why standard fairness metrics can conflict, using the real-world example of a lending model that passed statistical parity tests but failed an individual fairness audit. They discuss the Impossibility Theorem of Fairness, how researchers at MIT measured bias in commercial facial recogni...

When there's not enough real data to train a model, data scientists are turning to synthetic data — artificial datasets generated from a small sample of real observations. In this episode, Lucas and Luna unpack how a healthcare startup used synthetic data to train a rare-disease diagnostic model when only 200 real patient records existed. They walk through the generation techniques — from simple b...

In Episode 33 of The Data Science Podcast, Lucas and Luna drill into a single question that has vexed marketers and analysts alike for decades: how do you know if an ad really caused a sale, or if the person would have bought anyway? Rather than a theoretical overview, the episode centers on a concrete case from a mid-size e-commerce company that ran a geo-level experiment poking holes in its own...

Episode 32 of The Data Science Podcast explores the critical difference between deploying a model and keeping it relevant. Lucas and Luna break down why most ML teams treat model retraining as a fire drill instead of a scheduled process. They walk through the real-world case of an e-commerce company whose recommendation engine silently deteriorated over six months because no one monitored feature...

Episode 31 of The Data Science Podcast with Fexingo examines a growing crisis in applied AI: models that perform perfectly in testing but cause real-world harm. Lucas and Luna dissect a 2025 case where a major hospital network's sepsis prediction algorithm silently amplified racial disparities for two years before a junior data scientist flagged it. They walk through what went wrong — biased train...

Episode 30 dives into the documented case of a 2019 health-care algorithm used by millions of US patients that systematically assigned lower risk scores to Black patients than equally sick white patients. Lucas and Luna walk through how the model's training label — historical health-care spending — encoded structural inequality, why a team at the University of California Berkeley caught it years l...

Lucas and Luna dive into the hidden bias that plagued credit scoring models for decades—and the data science fix that's finally changing it. They walk through a 2025 study from the Consumer Financial Protection Bureau showing that traditional credit scores systematically under-score borrowers who pay rent on time but lack deep credit history. Using actual FICO and VantageScore data, Lucas explains...