Doug Safreno’s Post

After the recent OpenAI announcement of `gpt-4o-mini`, i'm sure we're all thinking about it : the cheap and fast option just got a whole lot smarter ! ♻️ Does that mean you should switch from `gpt-3.5` ? 👉 Well probably, but first you need to check how this new model performs with your existing tasks. Hugues breaks down for you how to use Literal AI to evaluate and validate your app's performance on different models.

Retrieval Augmented Generation (RAG) allows you to fine-tune LLMs by feeding them information they didn’t have when they were first trained. Learn to build a multi-file format RAG web app using Gradio, Groq, and LlamaIndex. - Thomas Reid https://hubs.la/Q02VVP_v0

Today's links look at advice for integrating LLMs into your app, why to keep blogging after many years, and what the semantic router pattern is all about in agentic workflows.

Let's face it, virtually every company on the internet is competing for attention with a couple of giant platforms. But there is hope: You can build the same kind of "For You Page" that everyone is addicted to on their favorite apps, but for *your* site or app, and we'll give you all the tools to make it super easy. Here's an explainer that even gets into all the technical details. https://lnkd.in/eN4N2tG2

Do you like Llama's? do you like tools? if you like both even better! 😅 New blog post out about using the new Llama 3.1 built-in tools! like video better? don't miss our youtube launch video here instead: https://lnkd.in/gt3njSuR

Vector Search 🔎 ✔️ What is it? ✔️ What are the benefits? ✔️ Differences between traditional and vector search? ✔️ What are the steps to building a vector search app? ✔️ What does its future look like? Check out our complete guide! 📕 https://bit.ly/4bVOJK9

Vector Search 🔎 ✔️ What is it? ✔️ What are the benefits? ✔️ Differences between traditional and vector search? ✔️ What are the steps to building a vector search app? ✔️ What does its future look like? Check out our complete guide! 📗 https://bit.ly/4bVOJK9

Useful intro to vector search!

🧑💻 Want to integrate multiple LLMs into your app? Try Portkey, your control panel for AI apps. Follow our step-by-step guide to learn how to get started routing requests to 200+ LLMs for vision, audio, and image generation. 🚀 https://lnkd.in/dmUxBapH

Hey LinkedIn👋 I'm excited to share my end-to-end sentiment analysis project of Medium app reviews from the Google Play Store! It can classify reviews as positive, negative, or neutral based on their text content. The dataset was extracted from Kaggle. Link - https://lnkd.in/gna7DQqs GitHub- https://lnkd.in/gFZyPHuU Here's what I did: - Exploratory data analysis: After a basic EDA I found out that there are no missing values & no null values. But there's a HUGE class imbalance. - Preprocessing: I did not simply remove data points to handle class imbalance as it leads to data loss which can hinder in the training process. Next, I removed numbers, special characters, stop words, etc. since they don't contribute to this use case. - Modeling preparation: Label encoded sentiment column & TF-IDF vectorized the reviews column. - Model building: Initially, trained a basic Random Forest Classifier and later hyperparameter tuned it using GridSearchCV. - Evaluation: Got a macro F1 score of 0.75 and an accuracy of 87%. It considers the performance of the model on both the majority and minority classes providing a better picture of its effectiveness. - Deployment: The app is hosted on Streamlit cloud, allowing users to easily predict the sentiment of the text they input. Looking forward to exploring further applications of this in the future! #NLP #machinelearning #datascience #data