Project
Google like Autocomplete - Platform-Level Suggester Microservice
I. Project Overview and Context
As Senior Software Engineer at Apna, a major job-hunting platform serving over 30 million app users for searching and applying for jobs, I designed and developed a platform-level Suggester microservice that made job search fast and relevant. The challenge was to enhance the job search system to be fast, relevant, and highly scalable.
II. Technical Architecture and High-Volume Scaling
The microservice was built using Go, Elasticsearch, and Redis. The service was designed and scaled to serve over 1,500 requests per second, achieving low latency with query responses consistently well below ten milliseconds. I implemented a mechanism to ingest search terms from the existing monolith system into Elasticsearch via Google Pub/Sub, capturing all updates and delete events.
III. Key Features and Conceptual Development
I conceptualized and implemented a search term interpretation layer crucial for handling popular, natural-language search queries. This layer supports complex search terms that include both filters and keywords (e.g., "WFH jobs in Delhi," "Part-time jobs in Bangalore"). I devised a wrapper language that allowed internal Operations teams or anyone to query Elasticsearch directly and map the queries to any search terms. Additionally, I headed a process for identifying duplicates within a corpus of three million search terms via a semi-automated process.
IV. Impact and Conclusion
The microservice contributed significantly to the platform's job search system, which is relied upon by millions of users. This project demonstrated expertise in designing highly scalable, low-latency microservices using specialized indexing and caching technologies (Go, Elasticsearch, Redis), and translating product needs (relevant search, supporting complex queries) into robust technical architectures.