Towards Data Science

Your time series model is failing to capture subtle local trends? Nikhil D.'s guide explains LOESS smoothing, showing how a weighted local regression adapts to data, smoothing out sharp spikes and dips for a more accurate trend line.

How do you evaluate an AI agent that calls tools and performs a multi-step task? Ida Silfverskiöld’s article on Agentic Evals explains how to measure Task Completion and Tool Correctness, metrics unique to agentic workflows.

Calling all Prompt Engineers! 🚀 Daphne de Klerk joins TDS, bringing her project management expertise to a nuanced guide on avoiding prompt bias. Her article is packed with insights on how a single word can steer an LLM. Inspired to contribute your own insights on AI or ML? We're a community-driven platform ready to publish your article: https://lnkd.in/gw8MqkPb

Uncover the secrets of LLM self-improvement. This article by Sudheer Singh reveals how AuPair leverages iterative code repairs to supercharge model learning, adding partially correct fixes back into the training data for continuous enhancement. Learn how this innovative approach could revolutionize the way we train AI.

Struggling with channel-level correlations in your CNN? This guide by Muhammad Ardi Putra unlocks the secrets of channel-wise attention, offering a from-scratch PyTorch implementation of the SE module. Level up your model's predictive power and leave those frustrating accuracy plateaus behind.

Unlock your competitive edge in the AI era. Marina Tosic's article shows non-technical professionals how to master AI skills, from high-level LLM understanding to evaluating outputs & building custom agents.

What's the RIGHT way to benchmark multi-step reasoning in LLM agents? Tomaz Bratanic dives into creating evaluation datasets that reflect how agents actually work with graph databases, beyond simple text-to-query translation.

Are you looking to optimize your ML model maintenance? Shafeeq Ur Rahaman teaches you to move beyond the "retrain reflex" by focusing on root cause analysis, fixing feature logic, and leveraging segment awareness for robust performance.

How will AI shape the Product Manager role? Rahul Vir dives into the Human-Centric Automation Matrix to help PMs prioritize tasks for AI automation vs. augmentation, focusing on worker desire and technical capability.

Can your API teach an AI agent what to do next? Roni Dover explores Conversation-Driven APIs for LLMs, demonstrating how contextual suggestions in responses lead to more effective multi-step investigations.