Strategies for Personalizing AI Recommendations

🤩 What if all your LLM interactions could be used to create a personalized layer that continuously evolves to adapt to your style? You wouldn’t need to keep reminding the model to match your style—it would just adapt naturally over time. Since we’re already giving LLMs so much data, why not use it to create a user-specific layer? Here's some good work in this direction! ⛳ A new paper proposes "PPlug", a lightweight plugin that creates a user-specific embedding based on historical user behaviors, allowing LLMs to tailor outputs without altering their structure. The plugin operates on a plug-and-play basis, meaning it improves personalization without retraining the model. It captures holistic user behavior rather than focusing on specific instances (generally done in RAG like approaches), leading to better adaptation to user preferences. It doesn't treat all user data equally, it selects relevant historical behaviors and synthesizes them into a personal embedding, based on their importance to the current task. 🤔 I really feel like personalization isn’t getting the attention it deserves, even though we have the technology to do it well. Probably because it’s quite expensive at this point, but it’s definitely something that will gain traction soon. Just imagine LLM-based personalization integrated into all AI products—it would completely change how we interact with tech. Link: https://lnkd.in/egWA_UZK

Intriguing insights from the Netflix Tech Blog (https://lnkd.in/gmd576ft) on their push towards a Foundation Model for personalized recommendations! Moving beyond numerous specialized models ("model first") to a unified, LLM-inspired system leveraging comprehensive user history ("data first") is a bold, potentially game-changing approach in the TMT space. Here are a few points that stood out for me: • Centralized learning: Share insights across every “Continue Watching” and “Today’s Top Picks.” • Smart tokenization: Merge clicks, plays, and scrolls into meaningful tokens. • Long‑term objectives: Predict multiple next interactions and genres, not just the next click. • Cold‑start: New shows get their moment, even before anyone’s watched them (highly intrigued by this). • Scale‑driven gains: Bigger data + bigger models = better recommendations. I see a few hurdles while scaling at the enterprise level, beyond the initial promise: 𝗧𝗵𝗲 𝗖𝗼𝗹𝗱 𝗦𝘁𝗮𝗿𝘁 𝗖𝗼𝗻𝘂𝗻𝗱𝗿𝘂𝗺: How effectively can a massive FM handle brand new content recommendations before interaction data exists? 𝗘𝘃𝗼𝗹𝘃𝗶𝗻𝗴 𝗪𝗼𝗿𝗹𝗱𝘀: Unlike static text corpus, media catalogs and user tastes change constantly. Can incremental training keep pace without breaking the bank or sacrificing relevance? 𝗦𝗰𝗮𝗹𝗲 & 𝗟𝗮𝘁𝗲𝗻𝗰𝘆: The sheer cost and complexity of training and serving inferences at Netflix scale, while maintaining low latency, sounds challenging. While consolidating models is appealing, the practicalities of cost, accuracy across diverse tasks, real-time adaptation, and managing ever-changing content libraries are critical considerations for any enterprise exploring this path. What are your thoughts? Is this the future of personalization, or are specialized models still king for specific use cases in the enterprise? #AI #FoundationModels #RecommendationSystems #Personalization #Netflix #EnterpriseAI #MachineLearning #TMT #Scalability #TechTrends #TMTatFractal #Fractal

Let's build a Recommender for an E-Commerce clothing site from scratch. 🛍️📈 This notebook shows how to deliver personalized, scalable recommendations even in cold-start scenarios. 👉 Product details include: - Price, - Rating, - Category, - Description, - Number of reviews, - Product name with brand. We have two user types, defined by their initial product choice at registration or general preferences around price range and review requirements. We'll use the Superlinked Framework to combine product and user data to deliver personalized recommendations at scale. Let's dive in 🏗️: 1️⃣ Data Preparation ⇒ Load and preprocess product and user data. 2️⃣ Set up the Recommender System ⇒ Define schemas for products, users, and user-product interactions. ⇒ Create embedding spaces for different data types to enable similarity retrieval. ⇒ Create the index, combining embedding spaces with adjustable weights to prioritize desired characteristics. 3️⃣ Cold-Start Recommendations ⇒ For new users without behavior data, we'll base recommendations on their initial product choice or general preferences, ensuring they're never left in the cold. 4️⃣ Incorporate User Behavior Data ⇒ Introduce user behavior data such as clicked, purchased, and added to the cart with weights indicating interest level. ⇒ Update the index to capture the effects of user behavior on text similarity spaces. 5️⃣ Personalized Recommendations ⇒ Now it's time to tailor recommendations based on user preferences and behavior data. ⇒ Compare personalized recommendations to cold-start recommendations to highlight the impact of behavior data. Ant that's a wrap! 🔁 Adjusting weights allows you to control the importance assigned to each characteristic in the final index. This tailors recommendations to desired behavior while keeping them fresh and relevant... it's easier than chasing the latest fashion trends. ✨ Dig into the notebook to implement this approach 👉 https://lnkd.in/edeQW344 Why not show some support by starring our repo? ⭐️ We'd appreciate it more than a free fashion consultation! 😉

Below is a diagram of our agentic architecture (well, part of it). See the top-right box: "recommender service"? Let’s talk about that. At Aampe, we split copy personalization into two distinct decisions: ➡️ Which item to recommend ➡️ How to compose the message that delivers it Each calls for a different approach. For item recommendations, we use classical recommender systems: collaborative filtering, content-based ranking, etc. These are built to handle high-cardinality action spaces — often tens or hundreds of thousands of items — by leveraging global similarity structures among users and items. For message personalization, we take a different route. Each user has a dedicated semantic-associative agent that composes messages modularly — choosing tone, value proposition, incentive type, product category, and call to action. These decisions use a variant of Thompson sampling, with beta distributions derived from each user’s response history. Why split the system this way? Sometimes you want to send content without recommending an item — having two separate processes makes that easier. But there are deeper reasons why recommender systems suit item selection and reinforcement learning suits copy composition: 1️⃣ Cardinality. The item space is vast — trial-and-error is inefficient. Recommenders generalize across users/items. Copy has a smaller, more personal space where direct exploration works well. 2️⃣ Objectives. Item recommendations aim at discovery — surfacing new or long-tail content. Copy is about resonance — hitting the right tone based on past response. 3️⃣ Decision structure. Item selection is often a single decision. Copy is modular — interdependent parts that must cohere. Perfect for RL over structured actions. 4️⃣ Hidden dimensions. Item preferences stem from stable traits like taste or relevance. Copy preferences shift quickly and depend on context — ideal for RL’s recency-weighted learning. 5️⃣ Reward density. Item responses are sparse. Every content delivery yields feedback — dense enough to train RL agents, if interpreted correctly. In short: recommenders find cross-user/item patterns in large spaces. RL adapts to each user in real time over structured choices. Aampe uses both — each matched to the decision it’s best for.