Katie Yu Xia

Food is more than just nourishment; it's deeply intertwined with our health, lifestyle, and culture. In the age of AI, eCommerce search is evolving from a tool into a source of inspiration. It's no longer just about showing you what you're looking for—it also offers delightful surprises you didn't know you wanted. Often, we don't know exactly what to add to our cart. Maybe we have a vague idea, a theme, or an impulse—craving traditional Asian cuisine, trying making sushi at home but needing guidance on the ingredients, planning a fun football night, aiming for a healthier lifestyle with gluten-free meals, or simply wondering, "What's for dinner?" Thanks to LLMs and their vast knowledge of brands, cuisines, recipes, nutrition, cultures, and more, we're making these experiences possible. It takes a dedicated team to seamlessly blend the world knowledge of LLMs with Instacart's insight into our users and product catalog. This involves careful engineering, thoughtful product design, and rigorous A/B testing with real-world traffic. It's been an exciting journey collaborating with such a talented team to supercharge eCommerce search with LLMs.

6

You want to do Machine Learning on large scale Marketing solutions ? Check below 🚀

3

More opportunities

12

🚀 I'm excited to share that our paper, "Enhancing E-commerce Product Title Translation with Retrieval-Augmented Generation and Large Language Models," has been accepted at CIKM 24' - Data-centric AI. Thanks for the co-authors Stephan Walter and Taichi Nakatani for the collaboration! 😀 Our study explores how we can improve the machine translation of e-commerce product titles using LLMs, particularly languages that the LLMs do not primarily support. The results of our experiments show that our proposed approach leverages large volumes of bilingual product data, and utilizes a probabilistic retrieval model to retrieve similar bilingual product information for few-shot learning, can improve overall product title translation quality by up to +15.3% in chrF scores. Notably, we observe that language pairs where the LLM has limited proficiency can benefit the most from our method, addressing a common challenge in this domain. As new and more capable LLMs continue to emerge, our RAG-based approach allows for the quick integration of these models (without extensive fine-tuning). By using the large volume of bilingual product data, including titles, bullet points, and descriptions, as few-shot learning examples, our method provides a flexible and adaptable solution that is particularly suitable for industrial applications, where the ability to keep pace with changing requirements is crucial. #amazonscience #LLM #CIKM #RAG #machinetranslation #searchengine #ecommerce #translation #GenerativeAI #search #MultilingualSearch #mixtral #localization #retail

65

2 Comments

Sharing my latest article on the RAFT: Teach LLMs to be better at RAG Discover how RAFT bridges the gap between fine-tuning and RAG, enhancing model performance by integrating domain knowledge with powerful retrieval techniques. 🔍 Dive into: • The mechanics of RAFT • Benefits over traditional methods • Real-world applications using Meta Llama 2 Read the full article here: https://lnkd.in/g7ZQNE5s #AI #MachineLearning #DataScience #NaturalLanguageProcessing #AIInnovation #MetaLlama2 #RAFT #llm #RAG #genAI #LargeLanguageModels

9

Resharing it: This is a great summary shared by the LinkedIn engineering team about their #llmops and LLM deployment work and their key learnings. #llms #chatgpt4 #gemini #aiml #mlops #llmops https://lnkd.in/d2bH5GRR

2

Hi LinkedIn Community,   Short version:  Grad student building low-code tool for ML data pipelines. Goal: Make ML dev work enjoyable again, not just endless debugging. Ask: Shadow sessions (see Calendly link below) Longer version: As ML devs, we often leave the office thinking, “Do I actually love doing my work?” because we had spent most of the day on inefficient whack-a-mole debugging and maintenance work. I want to get us back to where we can love coding for those reasons we originally chose to pick up coding in the first place.   A lot of AI tools are restrictive and end up mangling codebases, and I believe the current state of AI coupled with strong design principles can enable much better code native developer experiences. I’ve gone through prototypes/testing and have some competing hypotheses. I’d love to observe coders working on ML data pipelines (happy to sign NDAs if necessary, can also be personal projects instead). If you're open to sharing your expertise, please schedule a shadow session: https://lnkd.in/eBW6p6qE Let's make ML data development more efficient and actually enjoyable! #MachineLearning #DataPipelines #LowCode #DevTools #AI #SoftwareEngineering #GradStudent #Research #MLOps #DataScience #ArtificialIntelligence #ProductivityTools #DeveloperExperience #CodeQuality #TechInnovation #ComputerScience #DataEngineering #CloudComputing

84

8 Comments

For the past few weeks, I’ve been exploring JAX and Flax, and I’m really appreciating their design philosophy. A key decision they’ve made is to avoid “reinventing the wheel.” Rather than duplicating functions like dataset loading or checkpointing, JAX/Flax taps into the strengths of PyTorch and TensorFlow, which already offer robust solutions. With a few adjustments, you can seamlessly use these libraries within the JAX/Flax ecosystem for data handling, saving both time and resources. For checkpointing, Flax initially had its own system with `flax.training.checkpoints`, but this has since been replaced by Orbax, which enhances model persistence. However, recent changes in JAX's random number generation (RNG) format have introduced some challenges. The new random key format `dtype=key<fry>` is fully compatible with Flax itself, but Orbax has yet to adopt it, which means that saving checkpoints with dropout layers now requires additional manual adjustments to the key type, adding a layer of complexity. This experience highlights a broader challenge for Flax: its reliance on external libraries that may not always keep pace with its own updates. While the Flax community is actively innovating, dependency on other libraries means they’re sometimes at the mercy of external update cycles. It’s a reminder that in fast-evolving ecosystems, dependencies can sometimes introduce bottlenecks that are outside the control of the primary library's community. Despite these challenges, it’s exciting to see how JAX and Flax continue to push the boundaries, even if it means navigating some temporary friction along the way. #MachineLearning #DeepLearning #JAX #Flax #DataScience #TensorFlow #PyTorch #DeepLearning

12

🚀 Spend 15 minutes reading this article: "DoorDash's real-time processing system" By Vu Trinh👨‍💻 In this insightful article, Vu Trinh, a data engineer, takes us through how DoorDash built a scalable and fault-tolerant real-time event processing system using Apache Kafka and Apache Flink. 🛠️ 💡 Key takeaways: Transitioning from inefficient legacy pipelines to a scalable solution with Kafka & Flink. 🔄 Using Kafka's REST Proxy to simplify message production and boost throughput. 🚚 Event processing with Flink based on real-time, event-time logic, and integrating with Snowflake for data warehousing. 🧳 Implementing schema enforcement and validation for consistent data flow. ✅ This article offers a deep dive into DoorDash’s architecture, sharing practical insights for data engineers looking to build high-performance real-time systems. 🔍💻 👉 Read the full article https://lnkd.in/dsCHwS3A

2

Introducing TL;DL (Too Long Didn't Listen), a simple application that helps you engage with long-form podcasts more efficiently. As data scientists, Lawrence Wu and myself, often struggle to find the time to listen to multiple hours-long podcasts each week. To address this challenge, we developed TL;DL to leverage LLM summarization and provide us with concise summaries of these podcasts. The application automatically emails us summaries of new podcast episodes once their transcripts become available. This allows us to stay informed about the latest content without having to dedicate hours to listening to podcasts. And for those particularly lengthy episodes, like Lex Friedman's latest 4-hour podcast, TL;DL provides a convenient solution. In addition to email summaries, TL;DL also features a simple RAG (Retrieval Augmented Generation) application. We have vectorized over 1,000 podcasts, enabling us to ask questions and receive relevant answers based on the podcasts' content. While the current version lacks time and entity filters, it still surfaces interesting and informative results. For now, TL;DL offers two primary functions: subscription to email summaries of podcasts shortly after their release and RAG Q&A across podcasts. We acknowledge that the UI may require some improvement, but we hope that you find TL;DL as useful as we do. Our goal is to enhance its functionality and expand the content library over time. Give TL;DL a try at https://tldl.dev and let us know what you think!

30

2 Comments

I just wanted to share the profiles of a few great folks from the Northwestern University Master's in Data Science program. If anyone in my network is looking to hire some super smart people for your data teams, I highly recommend Lijia Yu, Yukang Wu, and Xinyi Wei!

23

3 Comments

Check out our latest blog to see how our Notification Team harnesses the power of Graph Neural Network (GNN) models to create personalized restaurant recommendations for user notifications.

63

1 Comment

I’m thrilled to collaborate with Josep from DataBites to share a comprehensive guide of top resources in ML. Josep Ferrer is a veteran technical writer of machine learning having published more than 50+ articles in Kdnuggets and towardsDataScience. Together, we've curated an essential list of resources to help you build a strong foundation in ML. The guide is divided into two parts. In Part 1, we share key theoretical resources: • Top-rated online courses • Must-read books. • Popular podcasts and tutorials to get you started. These high-quality resources are designed to equip you with the knowledge and confidence to tackle the challenges ahead in your ML journey. Checkout the article here: https://lnkd.in/gVDS_Uaw Stay tuned for Part 2 on Saturday, where we'll dive into practical, hands-on learning resources, some classic projects, and advanced techniques to help you apply your knowledge and build a standout portfolio. Don't miss out! Check out Part 1 now and take the first step. -- 𝘐𝘧 𝘺𝘰𝘶 𝘭𝘪𝘬𝘦𝘥 𝘵𝘩𝘦 𝘢𝘳𝘵𝘪𝘤𝘭𝘦, 𝘢𝘭𝘴𝘰 𝘤𝘰𝘯𝘴𝘪𝘥𝘦𝘳 𝘴𝘶𝘣𝘴𝘤𝘳𝘪𝘣𝘪𝘯𝘨, 𝑙𝑖𝑛𝑘 𝑏𝑒𝑙𝑜𝑤 𝑚𝑦 𝑝𝑟𝑜𝑓𝑖𝑙𝑒 𝑝𝑖𝑐𝑡𝑢𝑟𝑒 ☝️ #MachineLearning #LearningJourney #MLResources

69

20 Comments

𝐇𝐨𝐰 𝐭𝐨 𝐩𝐫𝐞𝐩𝐚𝐫𝐞 𝐟𝐨𝐫 𝐍𝐋𝐏 𝐢𝐧𝐭𝐞𝐫𝐯𝐢𝐞𝐰𝐬 𝐢𝐧 2 𝐦𝐨𝐧𝐭𝐡𝐬? The NLP/GenAI field has advanced significantly in recent years. Before applying for ML roles, I decided that I needed to catch up with the recent progress and refresh my knowledge. There were a few options to choose from, so I decided to go with the ones that were free and flexible in schedule. Below materials helped me to prepare and clear multiple ML interview rounds for GenAI roles across FAANG. I did interviews for ML Eng, Data Scientist and Research Eng positions (roles which didn’t require PhD). My main NLP preparation materials: 1. 💕 Stanford CS224N NLP with Deep Learning. The course recordings are available on YouTube - https://lnkd.in/eHSJxDsu . While some lectures were from 2021, new ones are being uploaded. The course starts from the basics (such as word2vec) and progresses to transformers. There aren't any specific topics that are asked more frequently than others, except for transformer architecture. So, it's good to understand all fundamentals like language models, word2vec, RNNs, LSTMs, exploding/vanishing gradients, BERT, GPT, and so on. 2. 💕 Lena Voita’s NLP Course  In order to get some hands-on practice, I was doing homework exercises from Lena’s open source repo - https://lnkd.in/eDjdKBV3 . Homeworks aligned nicely with Stanford’s lectures, it was a good addition.  You can also check out Lena’s website - https://lnkd.in/e7-_bCRV 3. 💕 Practice websites with questions: - https://lnkd.in/eEF2vxXD - https://lnkd.in/encKSQxx Is it enough? You also need to prepare for Computer Vision… I quickly learned that it is also very common to be asked Computer Vision and CNN questions while applying for NLP roles. You need to be familiar with concepts such as - max/min/avg pooling, kernel, padding, stride, 1D/2D convolution, in/out channels. 1. 🍀 I watched the Stanford’s recordings from 6 years ago (the course felt a bit outdated): https://lnkd.in/e8mpw2UB 2. 🍀 Searched for good medium and towardsdatascience articles like: Pytorch [Basics] — Intro to CNN: https://lnkd.in/eFyYHpkD 𝐼𝑓 𝑦𝑜𝑢 𝑘𝑛𝑜𝑤 𝑜𝑓 𝑎𝑛𝑦 𝑔𝑜𝑜𝑑 𝐶𝑜𝑚𝑝𝑢𝑡𝑒𝑟 𝑉𝑖𝑠𝑖𝑜𝑛 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙𝑠, 𝑝𝑙𝑒𝑎𝑠𝑒 𝑠ℎ𝑎𝑟𝑒 𝑡ℎ𝑒𝑚 𝑖𝑛 𝑡ℎ𝑒 𝑐𝑜𝑚𝑚𝑒𝑛𝑡𝑠. If you want to take one thing away from this post, you need to know the transformers architecture while interviewing in 2024. It became a routine for me to give a ~5min summary of it after the third time. You need to be able to dive deep into specific formula details as part of follow-up questions. I would suggest revising the architecture before every interview. Good luck with your interviews! 

32

Three insights from the TEDAI San Francisco X Factory X AI at Meta CodeGen Hackathon. Had the pleasure of judging a fantastic hackathon in SF yesterday. Here's the biggest things I took away from iteration on the cutting edge of GenAI for code: 1) 𝙂𝙚𝙣𝘼𝙄 𝙘𝙤𝙙𝙞𝙣𝙜 𝙨𝙩𝙖𝙧𝙩𝙚𝙙 𝙬𝙞𝙩𝙝 𝙖𝙘𝙩𝙞𝙫𝙚𝙡𝙮 𝙖𝙨𝙨𝙞𝙨𝙩𝙞𝙣𝙜 𝙙𝙚𝙫𝙨 𝙖𝙩 𝚠̲𝚛̲𝚒̲𝚝̲𝚎̲ 𝙩𝙞𝙢𝙚, 𝙗𝙪𝙩 𝙂𝙚𝙣𝘼𝙄 𝙖𝙜𝙚𝙣𝙩𝙨 𝙣𝙖𝙩𝙪𝙧𝙖𝙡𝙡𝙮 𝙛𝙞𝙩 𝙖𝙩 𝙙𝙞𝙛𝙛 𝚛̲𝚎̲𝚟̲𝚒̲𝚎̲𝚠̲ 𝙩𝙞𝙢𝙚. Iterative agenic workflows can run at review with automated vetting of generations, and only serve edit suggestions shown to improve a tested outcome (i.e. memory efficiency). Bonus is that this can scale with inference time compute to try more or less variations or revision cycles. 2) '𝘾𝙤𝙙𝙚' 𝙖𝙨 𝙖𝙣 𝙖𝙗𝙨𝙩𝙧𝙖𝙘𝙩 𝙘𝙤𝙣𝙘𝙚𝙥𝙩 𝙤𝙥𝙚𝙣𝙨 𝙖 𝙬𝙤𝙧𝙡𝙙 𝙤𝙛 𝙥𝙤𝙨𝙨𝙞𝙗𝙞𝙡𝙞𝙩𝙞𝙚𝙨 𝙛𝙤𝙧 𝙣𝙤𝙣-𝙨𝙤𝙛𝙩𝙬𝙖𝙧𝙚-𝙙𝙚𝙫 𝙚𝙣𝙜𝙞𝙣𝙚𝙚𝙧𝙞𝙣𝙜 𝙙𝙞𝙨𝙘𝙞𝙥𝙡𝙞𝙣𝙚𝙨. Many tools in these areas involve cumbersome UI's that ultimately translate points and clicks into complex config files, and LLM's are pretty good at editing these underlying files directly. 3) 𝘾𝙤𝙙𝙞𝙣𝙜 𝙖𝙗𝙞𝙡𝙞𝙩𝙞𝙚𝙨 𝙞𝙣 𝙢𝙤𝙙𝙚𝙡𝙨 𝙖𝙧𝙚 𝙜𝙚𝙩𝙩𝙞𝙣𝙜 𝙢𝙤𝙧𝙚 𝙘𝙤𝙢𝙥𝙖𝙘𝙩 𝙩𝙝𝙖𝙣 𝙚𝙫𝙚𝙧. One hackathon team was able to use a combination of synthetic data creation/refinement and model distillation (in this case from my near and dear Llama 405B) to fit highly advanced code translation ability in a 14B model. I have a feeling this trend will continue, which will open up more inference compute heavy coding opportunities like broad repo summarization. Bonus here - the resulting model was open sourced by Jason Lu: https://lnkd.in/gswMbCR3 Shout out to SHACK15 and Cerebral Valley for hosting. Incredible views and vibes.

19

3 Comments

LLMs are transforming the recommendation industry as we speak! From what Spotify plays next to the trending posts on your TikTok feed, LLMs are making their way on every stage of a typical modern recommendation system. This includes: 1️⃣ Feature Processing 2️⃣ Candidate Generation 3️⃣ Ranking 4️⃣ User Interaction I've just published an article exploring the 𝗹𝗮𝘁𝗲𝘀𝘁 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗼𝗻 𝗵𝗼𝘄 𝗟𝗟𝗠𝘀 𝗮𝗿𝗲 𝗽𝗼𝘄𝗲𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗿𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 behind the services you love and use. 🔗 Check it out here: https://lnkd.in/gcgiG4H7 Let’s discuss how we can leverage these innovations to transform our industries. Share your thoughts in the comments! As always, happy learning! -- 𝘐𝘧 𝘺𝘰𝘶 𝘭𝘪𝘬𝘦𝘥 𝘵𝘩𝘦 𝘢𝘳𝘵𝘪𝘤𝘭𝘦, 𝘢𝘭𝘴𝘰 𝘤𝘰𝘯𝘴𝘪𝘥𝘦𝘳 𝘴𝘶𝘣𝘴𝘤𝘳𝘪𝘣𝘪𝘯𝘨, 𝑙𝑖𝑛𝑘 𝑏𝑒𝑙𝑜𝑤 𝑚𝑦 𝑝𝑟𝑜𝑓𝑖𝑙𝑒 𝑝𝑖𝑐𝑡𝑢𝑟𝑒 ☝️ #MachineLearning #LLMs #RecommendationSystems

98

20 Comments

Life as an analytics engineer on Nathan's team is never dull! I highly recommend his team as a great place to work These are some real questions that my team and others have asked them to solve: * what are users' ratings of our transcript quality by region? * do we have users who are probably speaking French but getting transcribed as English? * how many transcribed contact center calls could be processed in one chunk using our LLM (e.g. to generate a summary)? * how often do customers call a contact center and then call back again within 7 days?

9

I just published Vol. 80 of "Top Information Retrieval Papers of the Week" on Substack. My Substack newsletter features the 7-10 most notable research papers on information retrieval (including recommender systems, search & ranking, etc.) from each week, with a brief summary, and links to the paper/codebase. This week’s newsletter highlights the following research work: 📚 Comparing Commercial and Open-Source LLMs for Systematic Review Query Generation, from TU Wien 📚 A Systematic Comparison of Ranking Algorithms for LLM Evaluation, from the University of Michigan 📚 A Comprehensive Analysis of LLM Representations for Regression Tasks, from Google 📚 Benchmarking Multi-modal Retrieval-Augmented Generation for Web Content, from Ma et al. 📚 Efficient Large Language Model Scaling for Industrial Dense Retrieval, from Huang et al. 📚 A Survey on LLM-as-a-Judge, from Gu et al. 📚 Reproducing and Adapting 2D Matryoshka Training for Information Retrieval, from Wang et al. 📚 A Fast and Exact Similarity Search Engine for Billion-Scale Data Series, from Schafer et al. 📚 A Review of Visual Question Answering in the Era of Large Language Models, from SYSU 📚 Integrating ID Embeddings and Language Models for Enhanced Sequential Recommendations, from Huawei #InformationRetrieval #ResearchPapers #CuratedContent #Newsletter  #substack

32

1 Comment

Optimizing Databricks LLM Pipelines with DSPy

42

3 Comments