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

Episode 103 of The Data Science Podcast with Fexingo. Lucas and Luna dive into Monte Carlo simulations — not as a textbook concept, but as a practical tool data scientists use to quantify uncertainty. They walk through a real-world case: a mid-size logistics company that used Monte Carlo to model delivery times under variable traffic, weather, and fuel costs. Lucas explains the math behind random...

In this episode, Lucas and Luna dive into the practical applications of Sentence-BERT (SBERT) for semantic search in production. They discuss how SBERT converts text into dense vector embeddings, enabling similarity search beyond keyword matching. The hosts walk through a real-world case study of a mid-sized e-commerce company that replaced its legacy Elasticsearch-based search with an SBERT-power...

Lucas and Luna explore the practical use of SHAP (SHapley Additive exPlanations) for interpreting complex machine learning models. They walk through a real-world example: a credit risk model from a mid-sized European fintech that needed regulatory compliance under GDPR. Lucas explains how SHAP values decompose a prediction into feature contributions, and why game theory provides a principled found...

For Episode 100 of The Data Science Podcast, Lucas and Luna explore the controversial practice of training AI models on synthetic data. With real-world examples from autonomous vehicle companies like Waymo and medical imaging startups, they discuss when synthetic data works, when it fails, and how the 'data cascade' problem threatens to pollute next-generation models. Lucas explains why some resea...

In this episode, Lucas and Luna dive into Temporal Fusion Transformers (TFT), a deep learning architecture that has changed how data scientists approach time series forecasting. They walk through a concrete case from a major European electricity utility that used TFT to predict hourly load across 20,000 substations with unprecedented accuracy. You'll learn how TFT handles multiple time series simu...

In this episode, Lucas and Luna dive into how Spotify uses reinforcement learning to personalize playlists like Discover Weekly and Release Radar. They break down the multi-armed bandit problem, explore how Spotify balances exploration vs. exploitation to keep listeners engaged, and discuss the cold-start challenge for new users. Lucas explains why 'bandit' algorithms aren't one-size-fits-all and...

Episode 97 dives into counterfactual explanations — the 'what if' tools helping data scientists debug models and build stakeholder trust. Lucas and Luna walk through a concrete example: a credit-approval model that rejected a loan applicant, and how a counterfactual explanation revealed a single feature — years at current address — was the deciding factor. They discuss practical implementation usi...

In this episode, Lucas and Luna dive into multimodal zero-shot learning, the technique that lets AI models like CLIP recognize objects, scenes, and text across images without ever being explicitly trained on those combinations. They explore a concrete use case: a retail startup using a pretrained multimodal embedding model to automatically tag 10,000 new product photos per day with zero labeling c...

In Episode 95 of The Data Science Podcast with Fexingo, Lucas and Luna dive into a practical yet underappreciated technique: using k-nearest neighbors for anomaly detection. They kick off with a real-world story from a major credit card processor that flagged a series of fraudulent transactions by measuring distance to the nearest legitimate patterns. Lucas explains why distance-based methods can...

In episode 94 of The Data Science Podcast with Fexingo, Lucas and Luna explore how active learning cuts labeling costs by 80 percent while maintaining model accuracy. Using a concrete example from a medical imaging startup training a rare-disease classifier, they walk through uncertainty sampling, query strategies, and the human-in-the-loop workflow. They compare pool-based versus stream-based act...

Episode 93 of The Data Science Podcast with Fexingo dives into Thompson Sampling, a Bayesian approach to online experimentation that balances exploration and exploitation better than traditional A/B testing. Lucas and Luna walk through a concrete example from a real e-commerce site that ran a 50-variant landing page test — and how Thompson Sampling found the winner in half the time with 30% less t...

Episode 92 of The Data Science Podcast with Fexingo dives into embedded analytics—the practice of integrating dashboards, reports, and AI insights directly into customer-facing products. Lucas and Luna explore how companies like Notion and Canva use embedded analytics to drive product-led growth, reduce churn, and monetize data. They walk through a real case: a fictional B2B SaaS platform that cut...

Most marketing attribution models are correlational — they tell you what happened, not why. In this episode, Lucas and Luna break down how data scientists are using causal inference techniques, specifically double machine learning and instrumental variables, to measure the true incremental impact of ad spend. Using a real 2025 case from a mid-market e-commerce brand that ran geo-lift tests across...

In this episode, Lucas and Luna explore how knowledge graphs are supercharging retrieval-augmented generation (RAG) systems. They break down a concrete example: how a financial services firm used a knowledge graph built from SEC filings and earnings call transcripts to reduce hallucination in their Q&A chatbot by 40 percent. The hosts explain why flat vector search alone often fails, how graph tra...

Lucas and Luna explore how graph neural networks are transforming recommendation systems, using the example of Pinterest's PinSage model. They break down how GNNs capture relational data like user-item interactions to generate high-quality recommendations, discuss the challenges of scaling to billions of nodes, and compare GNN-based approaches to traditional collaborative filtering. The episode in...

Episode 88 of The Data Science Podcast dives into dimensionality reduction — but not for preprocessing. Lucas and Luna explore how data scientists use t-SNE and UMAP to visualize high-dimensional data, from customer segmentation to single-cell genomics. They discuss the trade-offs between global and local structure preservation, the risk of over-interpreting clusters, and why a 2D plot is never th...

In episode 87 of The Data Science Podcast, Lucas and Luna explore manifold learning—a powerful technique for dimensionality reduction that goes beyond PCA. They focus on t-SNE and UMAP, explaining how these algorithms preserve local and global structure in high-dimensional data. Using concrete examples from genomics and image datasets, they discuss when to use each method and common pitfalls like...

In this episode, Lucas and Luna explore the concept of the Pareto frontier, a powerful framework for multi-objective optimization in data science. Starting with a concrete example—a ride-hailing company balancing driver wait times against passenger fares—they illustrate how Pareto optimality helps teams make trade-offs in model tuning, resource allocation, and product decisions. They discuss real-...

Lucas and Luna dive into Neural Radiance Fields (NeRFs), a technique that has reshaped 3D reconstruction from 2D images. They walk through how NeRFs work at a high level—converting sparse photographs into continuous volumetric scene representations—and why this matters for industries like autonomous driving, cultural heritage preservation, and virtual production. The episode anchors on a concrete...

In this episode of The Data Science Podcast, Lucas and Luna explore how data scientists are using diffusion models — the technology behind tools like DALL-E and Stable Diffusion — for image generation. They break down the core idea of gradually denoising random pixels into coherent images, discuss training and inference costs, and contrast diffusion models with GANs and autoregressive models. Usin...

In this episode, Lucas and Luna explore how data scientists apply transfer learning to solve image classification problems with very little labeled data. They break down the concrete steps: taking a pre-trained model like ResNet-50 trained on ImageNet's 14 million images, freezing early layers, fine-tuning later layers on a new task with as few as 50 images per class. Lucas shares a case study fro...

Lucas and Luna dive into Bayesian A/B testing—why it's replacing frequentist methods for marketing experiments. They walk through a real case from a mid-size e-commerce company that used Bayesian inference to compare email subject lines, reaching a decision in half the time with clearer probability statements. The episode covers the core difference: instead of a p-value, you get a direct probabili...

Federated learning is reshaping how organisations train machine learning models on sensitive data without ever centralising it. In this episode, Lucas and Luna break down a real-world case: how a consortium of six European hospitals used federated learning to train a diagnostic model for rare paediatric cancers — achieving accuracy comparable to a centralised model while keeping each hospital's pa...

Episode 80 of The Data Science Podcast dives into Shapley values — a game-theoretic approach to explaining model predictions. Lucas walks through the core intuition: how Shapley values fairly distribute prediction contributions among features, using a concrete example from a credit approval model. Luna asks about the practical trade-offs, including computational cost with high-dimensional data. Th...

Episode 79 of The Data Science Podcast explores synthetic control — a causal inference method that estimates what would have happened to a treated unit if the intervention never occurred. Lucas and Luna break down a real-world case: how a ride-hailing company used synthetic control to measure the impact of a surge-pricing algorithm change on driver supply in Austin, Texas. They walk through buildi...