Google's Releases Gemini 2.0

Hey there 👋

We hope you're excited to discover what's new and trending in AI, ML, and data science.

Here is your 5-minute pulse...

print("News & Trends")

Google releases Gemini 2.0 lineup and adds access to ‘thinking’ AI models (5 min. read)

Google unveils Gemini 2.0 updates, featuring the Flash Thinking AI model for complex reasoning and a flagship Pro version with enhanced coding and math capabilities. Alongside, the budget-friendly Flash-Lite model debuts, as Google ramps up AI investments to outpace rivals like OpenAI and Microsoft.

ByteDance’s OmniHuman-1: Deepfakes Just Got Real (6 min. read)

OmniHuman-1 generates ultra-realistic deepfake videos from a single image + audio, handling cartoons, human poses, and motion edits. Trained on 19K hours of video, it blurs reality—regulation and detection lag behind as AI-generated content becomes indistinguishable from real footage.

Hugging Face Replicating OpenAI's Deep Research (9 min. read)

Hugging Face attempted to replicate OpenAI's Deep Research, an agentic web-search framework that significantly improved GAIA benchmark performance, by conducting a 24-hour experiment to open-source an equivalent system.

print("Applications & Insights")

Automated Feature Engineering in PyCaret
PyCaret streamlines machine learning by automating feature engineering tasks like handling missing data, encoding categorical variables, scaling features, and detecting outliers. This user-friendly approach saves time and enhances model performance, making it ideal for both beginners and experts.

Time Series Forecasting with MFLES (7 min. read)
MFLES is a new time series forecasting method that uses gradient-boosted decomposition, applying learning rates to components like trend and seasonality. It excels in handling multiple seasonalities and variable histories, offering improved accuracy over traditional models.

How to Fine-Tune DeepSeek-R1 for Your Custom Dataset (Step-by-Step)
Step-by-step guide to fine-tuning the DeepSeek-R1 language model using the Unsloth library. The tutorial covers installing necessary libraries, loading the model, applying Low-Rank Adaptation (LoRA) for efficient training, and preparing datasets, enabling customization even on hardware with limited resources.

3 Easy Ways to Create Flowcharts and Diagrams Using LLMs
LLMs can automate flowchart creation using Mermaid.js, PlantUML, and Graphviz. Describe your process in text, and AI converts it into structured diagrams. This approach simplifies workflow visualization, making it faster and more accessible—perfect for engineers, analysts, and anyone needing clear process diagrams without manual effort.

print("Lectures & Courses")

Use RAG to chat with PDFs using Deepseek, Langchain and Streamlit (Video)
Build a production-ready AI chatbot that answers PDF queries with DeepSeek’s LLM and LangChain. This step-by-step guide covers setting up DeepSeek’s advanced reasoning model, integrating it with LangChain’s RAG system, and deploying a fast, interactive Streamlit interface for seamless document interaction.

Deep Dive into LLMs like ChatGPT (Video)
OpenAI cofounder Andrej Karpathy just released a 3.5-hour YouTube deep dive on how LLMs are trained, breaking down their inner workings and offering mental models to understand their ‘psychology’—a must-watch for anyone looking to grasp the fundamentals of AI.

How Transformer LLMs work (Course)
This Deeplearning.ai course breaks down transformer LLMs, covering tokenization, attention mechanisms, and cutting-edge advancements like Mixture of Experts (MoE) and KV cache, giving you a solid foundation in modern AI architectures.

That’s it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve the pulse experience for you.

See you soon,

Andres