Building a Recommendation Engine for Bus Search Based on Real-Time Feedback of Users
summary
A ranking microservice built in Golang and gRPC uses a Kafka feedback loop, Redis sorted sets and Lua scripts to pull the last 6–7 hours of interaction signals and 3–10 days of transactions, builds 28 features per service, caches LightGBM models from S3 and scores inventory in-memory; parallel goroutines and a channel pool cut fetch time to sub‑millisecond and the system returns personalized results in 2.5ms over the network with service latency as low as 0.5ms at scale, validated by A/B lift in conversion.
AI-written summary · about
chapters (6)
- 0:00Introduction to bus search
- 3:13Learn to rank algorithm
- 10:07Data aggregation with Redis
- 15:06gRPC communication setup
- 18:16Performance and scaling results
- 18:57Q&A session
related talks
55mBuilding a Recommendation Engine with Machine Learning Techniques
01h 14mBuilding a Recommendation System with a Hybrid Search From Scratch
23mML-Powered Search & Recommendation 101
25mBetter search relevance using Learning to Rank at mobile.de
33mPowering Real-time Recommendations with Machine Learning on GCP
32mResurrecting a Recommendations Platform
27mUsing Graph Theory to Build a Recommendation Engine in Node js
20mDataEngBytes 2023 - PER-T1-06 - Kamal Hossain