Tecton

Batch data processing can create a critical ML performance plateau for travel marketplaces. ?? While competitors struggle with stale insights, HomeToGo now processes 100K requests/second with sub-second latency by integrating Snowplow's real-time data collection with Tecton The result? Hyper-personalized search without massive engineering overhead. Swipe below to learn more ?? See the full story: https://lnkd.in/eyhfMKte #RealTimeML #TravelTech

Paul Iusztin delivers again! This time with a comprehensive guide to architecting fraud detection solutions. He provides a complete view including raw data transformation, feature engineering for production, low latency serving/inferencing and the feedback loop to monitor feature and model drift. Check it out! https://lnkd.in/gnAQXj84

HomeToGo delivers personalized vacation rentals at scale by leveraging Snowplow and Tecton for powerful real-time data infrastructure. By capturing user interactions across their marketplace with 20 million+ vacation rental offers, HomeToGo is now: → Reducing feature freshness latency to sub-second speeds, enabling truly real-time personalization → Deploying advanced ML capabilities in just weeks with a small team of only two ML engineers → Processing 100,000+ requests per second with latencies below 100ms → Breaking through performance plateaus that limited their batch-only approach Learn how HomeToGo is creating competitive advantage through real-time personalization without massive engineering resources. Read the full story: https://lnkd.in/edqEDB8C #Snowplow #Tecton #DataInfrastructure #MachineLearning #Personalization #RealTimeData #TravelTec

Real-time machine learning is only as powerful as the data that fuels it. ?? At HomeToGo, ensuring our ranking algorithms deliver the most relevant and high-quality vacation rental options requires fresh, accurate, and fast-moving data. That’s why we’re always looking to optimize our feature freshness — a crucial but often overlooked challenge in real-time ML. Tecton breaks it down in their latest blog post by Alex Gnibus, leveraging insights from HomeToGo and Stephan, our Director of Data Analytics, to explore how organizations can combat feature staleness with the right architecture and engineering strategies. ?? Check out the full article at the link in our comments! ?? #hometogo #tecton #tech #data #insights #blog #innovation #ml

Time-sensitive ML applications like fraud detection and search ranking need extremely fresh features. Every millisecond in your feature pipeline matters. But how do you engineer a system where features go from raw event to prediction in milliseconds??See how HomeToGo did it in a matter of weeks using Snowplow + Tecton: https://lnkd.in/gaMhuMaT

Big news!?Tecton has been named one of?Forbes' 2025 America's Best Startup Employers—for the?third year in a row! https://lnkd.in/dGSX3ahr We’re honored to be ranked?#179 out of 500?among the top startups shaping the future. And guess what??We’re hiring!?If you’re passionate about?AI, machine learning, and innovation, check out our open roles. #Startups #AI #MachineLearning #BestStartupEmployers #CareerGrowth #HiringNow

Many ML projects fail to transition from POC to production-ready. Here's one simple reason why: The initial focus was never on scalability or production constraints. Fraud detection presents a perfect use case for building production-first ML systems. It combines the need for real-time and batch processing, low-latency predictions, and high-accuracy models. Here's a look into what that system could look like: ???????? ?????????????? In fraud detection, you deal with real-time transactions, streaming data and historical records. Real-time and streaming data require you to instantly compute features, while historical records help track user profiles and spending patterns. In our architecture, real-time features are computed through HTTP requests, streaming data will flow through Kafka topics, with historical data stored in a data warehouse for batch processing. ?????????????? ???????????????? At the heart of the system is the feature platform, like Tecton ... This centralizes all the features. More specifically, it allows us to manage features in an offline store (for training - high throughput) and online store (for serving - low latency). Using the same feature engineering logic during training and inference avoids training-serving skew. ?????????????? ?????????????????? These convert raw data into meaningful features. By centralizing your features into a feature store like Tecton, you can leverage their feature views to define features once and reuse them across models and pipelines. A feature view is defined as a data source(s) + a function that maps raw data into features. Next, using a process known as materialization, you sync the raw data sources with the online/offline stores while applying the transformations. ???????????????? ?????????????????? It ingests features and labels to train models (stored in a model registry). Leveraging the feature store, you can easily apply time-traveling strategies to version your dataset. ?????????????????? ?????????????????? It takes new transaction data, enriches it with features from Tecton's feature platform, and applies the trained model to generate predictions. Online stores are crucial at serving time. They provide low-latency access to up-to-date features. When a transaction occurs... The pipeline quickly retrieves pre-computed features, combines them with real-time features, and computes predictions. ?????????????????????????? ???????????????? Lastly, an observability pipeline is essential for monitoring the system's health and detecting drifts. The final touch is an alarm system that sends emails or SMS or denies transactions if fraud is detected. Want to dive deeper into building such systems? Check out the link in the comments.

? There's still time to catch our technical talk TOMORROW! ? Join AWS Solution Architect Arnab Sinha and Tecton's Isaac Cameron as we tackle one of ML's biggest headaches: months-long engineering cycles. Learn how Amazon SageMaker + Tecton can help you turn feature development from a lengthy process into minutes. Tecton's SageMaker integration simplifies the architecture you need for demanding, real-time use cases like fraud detection. Plus, we'll walk through a live demo so you can see it all in practice. Get your spot here!

How long did your last feature take to reach production? If your answer is "too long," join our technical deep-dive with AWS Solution Architect Arnab Sinha and Tecton's Isaac Cameron. They'll demonstrate a faster approach to ML deployment using Amazon SageMaker + Tecton. This live talk will include a live demo that can take you from notebook to training in minutes (instead of months). Register now! #MLOps #MachineLearning #MLEngineering https://lnkd.in/ghXcDDJ6

91% of ML models degrade in production. Your feature workflow might be the culprit causing accuracy issues: ?? Data leakage from complex feature dependencies ?? Inconsistent online/offline implementation ? Stale features that don’t reflect current patterns We've documented a framework that breaks down these challenges and the architectural patterns needed to improve model accuracy. Get the guide here! #MLOps #MachineLearning