Towards Data Science
@towardsdatascience.com
The world's leading publication for data science and artificial intelligence professionals.
Website 🌐 towardsdatascience.com
Submit an Article ✍️ https://contributor.insightmediagroup.io
Subscribe to our Newsletter 📩 https://bit.ly/TDS-Newsletter
Website 🌐 towardsdatascience.com
Submit an Article ✍️ https://contributor.insightmediagroup.io
Subscribe to our Newsletter 📩 https://bit.ly/TDS-Newsletter
How do AI agents decide what to do next? Kenneth Leung breaks down agentic planning in @langchain.bsky.social through the familiar idea of to-do lists.
How Agents Plan Tasks with To-Do Lists | Towards Data Science
Understanding the process behind agentic planning and task management in LangChain
towardsdatascience.com
January 17, 2026 at 4:27 PM
How do AI agents decide what to do next? Kenneth Leung breaks down agentic planning in @langchain.bsky.social through the familiar idea of to-do lists.
Transforming variables can quietly distort probabilities. Aniruddha Karajgi explains how the Jacobian adjustment keeps them honest.
Keeping Probabilities Honest: The Jacobian Adjustment | Towards Data Science
An intuitive explanation of transforming random variables correctly.
towardsdatascience.com
January 17, 2026 at 2:47 PM
Transforming variables can quietly distort probabilities. Aniruddha Karajgi explains how the Jacobian adjustment keeps them honest.
Learn how to build a maximum-efficiency coding setup. Eivind Kjosbakken's latest article walks through the tools, techniques, and mindset that can significantly increase your programming productivity.
Maximum-Effiency Coding Setup | Towards Data Science
Learn how to be a more efficient programmer
towardsdatascience.com
January 17, 2026 at 12:45 AM
Learn how to build a maximum-efficiency coding setup. Eivind Kjosbakken's latest article walks through the tools, techniques, and mindset that can significantly increase your programming productivity.
Running out of VRAM during the final step of LLM training? It's often the logit bottleneck. In a new article, Ryan Pégoud explains how to build a custom Triton kernel that fuses the linear and cross-entropy layers, cutting peak memory by 84%.
Cutting LLM Memory by 84%: A Deep Dive into Fused Kernels | Towards Data Science
Why your final LLM layer is OOMing and how to fix it with a custom Triton kernel.
towardsdatascience.com
January 16, 2026 at 10:05 PM
Running out of VRAM during the final step of LLM training? It's often the logit bottleneck. In a new article, Ryan Pégoud explains how to build a custom Triton kernel that fuses the linear and cross-entropy layers, cutting peak memory by 84%.
Remove the stubborn “yellow halo” from your AI image composites. Eric Chung's new article explains how to fix color contamination by switching from RGB to the Lab color space for a more precise, perceptual blend.
From RGB to Lab: Addressing Color Artifacts in AI Image Compositing | Towards Data Science
A multi-tier approach to segmentation, color correction, and domain-specific enhancement
towardsdatascience.com
January 16, 2026 at 8:30 PM
Remove the stubborn “yellow halo” from your AI image composites. Eric Chung's new article explains how to fix color contamination by switching from RGB to the Lab color space for a more precise, perceptual blend.
Understand why the growth of data giants like Databricks and Snowflake might be slowing down. Hugo Lu's new article explains the market forces and competitive pressures that suggest they are approaching a ceiling.
The Great Data Closure: Why Databricks and Snowflake Are Hitting Their Ceiling | Towards Data Science
Acquisitions, venture, and an increasingly competitive landscape all point to a market ceiling
towardsdatascience.com
January 16, 2026 at 7:18 PM
Understand why the growth of data giants like Databricks and Snowflake might be slowing down. Hugo Lu's new article explains the market forces and competitive pressures that suggest they are approaching a ceiling.
Welcome Suriyaa MM! 👋 In his debut article, Suriyaa explains how data-packing can achieve up to 3x speedups on memory-bound operations.
Submit your article to get published too 👉 bit.ly/TDSContributor
Submit your article to get published too 👉 bit.ly/TDSContributor
Breaking the Hardware Barrier: Software FP8 for Older GPUs | Towards Data Science
Deep learning workloads are increasingly memory-bound, with GPU cores sitting idle while waiting for data transfers. FP8 precision solves this on newer hardware, but what about the millions of RTX 30…
towardsdatascience.com
January 16, 2026 at 6:11 PM
Welcome Suriyaa MM! 👋 In his debut article, Suriyaa explains how data-packing can achieve up to 3x speedups on memory-bound operations.
Submit your article to get published too 👉 bit.ly/TDSContributor
Submit your article to get published too 👉 bit.ly/TDSContributor
Reposted by Towards Data Science
Beyond Prompting: The Power of Context Engineering | Towards Data Science towardsdatascience.com/beyond-promp...
Beyond Prompting: The Power of Context Engineering | Towards Data Science
Using ACE to create self-improving LLM workflows and structured playbooks
towardsdatascience.com
January 14, 2026 at 5:22 AM
Beyond Prompting: The Power of Context Engineering | Towards Data Science towardsdatascience.com/beyond-promp...
Reposted by Towards Data Science
If you're aiming for a data engineering career, this roadmap outlines exactly what to learn and in what order.
Data Engineer Roadmap
Step by step guide to becoming a Data Engineer in 2025
roadmap.sh
January 15, 2026 at 5:17 PM
If you're aiming for a data engineering career, this roadmap outlines exactly what to learn and in what order.
Reposted by Towards Data Science
How to Run Coding Agents in Parallel in @towardsdatascience.com
towardsdatascience.com/how-to-run-c...
towardsdatascience.com/how-to-run-c...
How to Run Coding Agents in Parallel | Towards Data Science
Get the most out of Claude Code
towardsdatascience.com
January 16, 2026 at 12:26 PM
How to Run Coding Agents in Parallel in @towardsdatascience.com
towardsdatascience.com/how-to-run-c...
towardsdatascience.com/how-to-run-c...
From chatbots to multi-agent systems, Mariya Mansurova walks through how the NeMo Agent Toolkit simplifies production-ready LLMs.
Production-Ready LLMs Made Simple with the NeMo Agent Toolkit | Towards Data Science
From simple chat to multi-agent reasoning and real-time REST APIs
towardsdatascience.com
January 16, 2026 at 4:47 PM
From chatbots to multi-agent systems, Mariya Mansurova walks through how the NeMo Agent Toolkit simplifies production-ready LLMs.
From fundamentals to transformers, this Advent Calendar reaches its final chapters. Angela Shi explores how modern NLP captures context, all within Excel.
The Machine Learning “Advent Calendar” Day 24: Transformers for Text in Excel | Towards Data Science
An intuitive, step-by-step look at how Transformers use self-attention to turn static word embeddings into contextual representations, illustrated with simple examples and an Excel-friendly…
towardsdatascience.com
January 16, 2026 at 3:03 PM
From fundamentals to transformers, this Advent Calendar reaches its final chapters. Angela Shi explores how modern NLP captures context, all within Excel.
Feel like you're just waiting around after giving a coding agent a task? That downtime is a productivity killer. Eivind Kjosbakken's new article dives into how to run multiple agents in parallel so you're always making progress.
How to Run Coding Agents in Parallel | Towards Data Science
Get the most out of Claude Code
towardsdatascience.com
January 16, 2026 at 5:02 AM
Feel like you're just waiting around after giving a coding agent a task? That downtime is a productivity killer. Eivind Kjosbakken's new article dives into how to run multiple agents in parallel so you're always making progress.
Learn how to build a data-driven vision board to track daily habits and long-term goals. Sabrine Bendimerad shares this new complete guide using #Python, Streamlit, and Neon.
The 2026 Goal Tracker: How I Built a Data-Driven Vision Board Using Python, Streamlit, and Neon | Towards Data Science
Designing a centralized system to track daily habits and long-term goals
towardsdatascience.com
January 15, 2026 at 10:05 PM
Learn how to build a data-driven vision board to track daily habits and long-term goals. Sabrine Bendimerad shares this new complete guide using #Python, Streamlit, and Neon.
Feeling like your team's code quality is inconsistent because it relies on people remembering standards? Erika Gomes-Gonçalves shows how automated guardrails like linters and tests can enforce standards and stop prototype issues from becoming production bugs.
Do You Smell That? Hidden Technical Debt in AI Development | Towards Data Science
Why speed without standards creates fragile AI products
towardsdatascience.com
January 15, 2026 at 7:18 PM
Feeling like your team's code quality is inconsistent because it relies on people remembering standards? Erika Gomes-Gonçalves shows how automated guardrails like linters and tests can enforce standards and stop prototype issues from becoming production bugs.
Agentic systems often give yes-or-no answers, but how do you evaluate them properly? Lambert Leong explains how to make AUC work for agentic AI without breaking evaluation standards.
Agents Under the Curve (AUC) | Towards Data Science
Towards understanding if your agentic solution is actually better
towardsdatascience.com
January 15, 2026 at 5:28 PM
Agentic systems often give yes-or-no answers, but how do you evaluate them properly? Lambert Leong explains how to make AUC work for agentic AI without breaking evaluation standards.
An LLM that’s 41× more efficient and 9× faster challenges the “bigger is better” mindset. Moulik Gupta breaks down how binary models rethink performance at scale.
What Happens When You Build an LLM Using Only 1s and 0s | Towards Data Science
An LLM that's 41× more efficient and 9× faster than today's standard models
towardsdatascience.com
January 15, 2026 at 3:03 PM
An LLM that’s 41× more efficient and 9× faster challenges the “bigger is better” mindset. Moulik Gupta breaks down how binary models rethink performance at scale.
Ever had an embarrassing moment from a simple pronunciation mistake? Samir Saci shares a story from a job interview and explains how it led him to build an AI pronunciation coach for Mandarin.
How AI Can Become Your Personal Language Tutor | Towards Data Science
How I used n8n to build AI study partners for learning Mandarin: vocabulary, listening, and pronunciation correction.
towardsdatascience.com
January 15, 2026 at 1:34 AM
Ever had an embarrassing moment from a simple pronunciation mistake? Samir Saci shares a story from a job interview and explains how it led him to build an AI pronunciation coach for Mandarin.
Is your file transfer strategy ready for 2026? Sign up for this free webinar from our sister site, @thenewstack.io, on Jan 29 to explore how AI automation and API-driven integration are creating secure, compliance-ready data exchanges for the modern enterprise.
bit.ly/fortra-mft-w...
bit.ly/fortra-mft-w...
January 14, 2026 at 11:52 PM
Is your file transfer strategy ready for 2026? Sign up for this free webinar from our sister site, @thenewstack.io, on Jan 29 to explore how AI automation and API-driven integration are creating secure, compliance-ready data exchanges for the modern enterprise.
bit.ly/fortra-mft-w...
bit.ly/fortra-mft-w...
Improve your generative AI applications with better context. Learn how knowledge graphs provide the structured data foundation needed for more reliable LLMs in @stevehedden.bsky.social's new article.
What Is a Knowledge Graph — and Why It Matters | Towards Data Science
How structured knowledge became healthcare’s quiet advantage
towardsdatascience.com
January 14, 2026 at 10:27 PM
Improve your generative AI applications with better context. Learn how knowledge graphs provide the structured data foundation needed for more reliable LLMs in @stevehedden.bsky.social's new article.
Are your dashboards created, applauded, and then quietly forgotten? Rashi Desai's new article explains how to design analysis around the decisions people actually need to make, not just the metrics.
Why Human-Centered Data Analytics Matters More Than Ever | Towards Data Science
From optimizing metrics to designing meaning: putting people back into data-driven decisions
towardsdatascience.com
January 14, 2026 at 9:15 PM
Are your dashboards created, applauded, and then quietly forgotten? Rashi Desai's new article explains how to design analysis around the decisions people actually need to make, not just the metrics.
Learn how to perform "surgery" on trained ViT models to eliminate artifacts. Jonathan Williford explains the "Test-Time Registers" method, a zero-retraining-cost solution to a common Transformer problem.
Glitches in the Attention Matrix | Towards Data Science
A history of Transformer artifacts and the latest research on how to fix them
towardsdatascience.com
January 14, 2026 at 8:15 PM
Learn how to perform "surgery" on trained ViT models to eliminate artifacts. Jonathan Williford explains the "Test-Time Registers" method, a zero-retraining-cost solution to a common Transformer problem.
Stop manually labeling topics. Discover how to use LLMs to automatically generate human-readable topic names and descriptions from your model's output in a this new article by Petr Koráb, Martin Feldkircher, and Márton Kardos.
Topic Modeling Techniques for 2026: Seeded Modeling, LLM Integration, and Data Summaries | Towards Data Science
Seeded topic modeling, integration with LLMs, and training on summarized data are the fresh parts of the NLP toolkit.
towardsdatascience.com
January 14, 2026 at 7:18 PM
Stop manually labeling topics. Discover how to use LLMs to automatically generate human-readable topic names and descriptions from your model's output in a this new article by Petr Koráb, Martin Feldkircher, and Márton Kardos.
Worried about your Text-to-SQL AI making mistakes? Even 90% accuracy can lead to bad business decisions and erode user trust. Gary Zavaleta's debut TDS article explains why the standard for enterprise AI must be binary: it works, or it's useless.
Why 90% Accuracy in Text-to-SQL is 100% Useless | Towards Data Science
The eternal promise of self-service analytics
towardsdatascience.com
January 14, 2026 at 6:17 PM
Worried about your Text-to-SQL AI making mistakes? Even 90% accuracy can lead to bad business decisions and erode user trust. Gary Zavaleta's debut TDS article explains why the standard for enterprise AI must be binary: it works, or it's useless.
See how RAG pipelines perform on real-world, messy questions versus clean, synthetic ones. Ida Silfverskiöld demonstrates why advanced features shine when dealing with vague or complex user queries.
When Does Adding Fancy RAG Features Work? | Towards Data Science
Looking at the performance of different pipelines
towardsdatascience.com
January 14, 2026 at 5:23 PM
See how RAG pipelines perform on real-world, messy questions versus clean, synthetic ones. Ida Silfverskiöld demonstrates why advanced features shine when dealing with vague or complex user queries.