Multi-Language Article Search
Search infrastructure serving 60M+ articles across 300 languages with sub-100ms response times. Handles 15,000+ QPS with language-specific tokenization, relevance tuning, and autocomplete.
Geospatial Driver Matching
Real-time geospatial search matching riders with nearby drivers. Indexes millions of location updates per second with geo-distance queries and multi-factor filtering.
Observability Log Search
Centralized logging infrastructure indexing petabytes of application logs. Enables engineers to search across thousands of microservices for debugging and incident response.
Code Search
Search across 200M+ repositories with code-aware indexing. Handles special characters, language-specific syntax, and filtering by repository metadata.
What Elasticsearch Engineers Actually Build
Before writing your job description, understand what Elasticsearch work looks like in production. Here are real examples from industry leaders:
Search & Discovery
Wikipedia uses Elasticsearch to power search across 60+ million articles in 300+ languages. Their search engineers handle:
- Multi-language tokenization and stemming (Japanese vs. English are very different)
- Relevance tuning so "Apple" returns the company or fruit based on context
- Search suggestions and autocomplete with sub-100ms response times
- Handling 15,000+ queries per second during peak traffic
GitHub uses Elasticsearch for code search across 200+ million repositories:
- Searching within code files (not just filenames)
- Filtering by language, repository, and organization
- Handling special characters in programming languages (regex, operators)
Geospatial & Real-Time
Uber uses Elasticsearch for geospatial search—matching riders with nearby drivers:
- Geo-distance queries across millions of driver locations
- Real-time indexing as drivers move (location updates every few seconds)
- Combining geo-queries with filters (driver rating, vehicle type)
Instacart uses similar patterns for store and product search:
- Finding stores within delivery radius
- Real-time inventory updates affecting search results
- Personalized product ranking based on purchase history
Logging & Observability (ELK Stack)
Netflix uses Elasticsearch as part of their observability stack:
- Indexing petabytes of application logs
- Enabling engineers to search error patterns across thousands of microservices
- Correlating logs with metrics and traces for debugging
Datadog competes with the ELK stack but also uses Elasticsearch internally:
- Time-series log aggregation
- Full-text search across structured and unstructured log data
- Alerting based on log patterns
E-Commerce
Zalando (Europe's largest fashion platform) uses Elasticsearch for:
- Product search with faceted filtering (size, color, brand, price)
- Personalized search rankings based on user behavior
- Synonym handling (sneakers = trainers = athletic shoes)
Skills by Level: Junior to Senior Progression
Junior Search Engineer (0-2 years)
What they can do:
- Write basic Elasticsearch queries using Query DSL
- Create simple mappings and index configurations
- Integrate search into a backend application
- Debug common issues (mapping conflicts, query syntax errors)
What they'll need help with:
- Performance optimization and cluster sizing
- Complex relevance tuning
- Handling cluster failures and recovery
- Designing schemas for scale
Mid-Level Search Engineer (2-4 years)
What they can do:
- Design index mappings optimized for specific query patterns
- Implement relevance tuning with boosting and function scores
- Set up monitoring and alerting for cluster health
- Handle reindexing and schema migrations
- Optimize queries that are hitting performance limits
What they'll need help with:
- Multi-cluster architectures
- Complex aggregation pipelines
- Capacity planning for rapid growth
Senior Search Engineer (4+ years)
What they can do:
- Architect multi-cluster, multi-region Elasticsearch deployments
- Design search relevance strategies from scratch
- Lead major version upgrades with zero downtime
- Build custom analyzers and plugins for specific use cases
- Make build-vs-buy decisions (Elasticsearch vs. Algolia vs. custom)
- Mentor team on search fundamentals and best practices
What distinguishes them:
- Experience with search at scale (millions of documents, thousands of QPS)
- Deep understanding of Lucene internals (not just Elasticsearch DSL)
- Ability to translate business requirements into search architecture
The Modern Elasticsearch Engineer (2024-2026)
Elasticsearch has evolved significantly since its early days. The ecosystem looks different now, and so do the required skills.
The Cloud Shift
Most production Elasticsearch deployments now run on Elastic Cloud, AWS OpenSearch, or managed Kubernetes clusters. Self-managed clusters are becoming rare except at the largest companies. This means:
- Less focus on bare-metal cluster administration
- More emphasis on cloud configuration, monitoring, and cost optimization
- Experience with Terraform/Pulumi for infrastructure-as-code
OpenSearch: The Fork You Need to Know About
In 2021, Elastic changed their license, and AWS forked Elasticsearch to create OpenSearch. Many companies, especially in regulated industries, now use OpenSearch instead. A strong Elasticsearch candidate should:
- Understand the Elasticsearch/OpenSearch split and its implications
- Be able to work with either (the APIs are nearly identical through version 7.x)
- Know which features diverged post-fork
Vector Search and AI Integration
Modern search engineers are expected to understand:
- Vector embeddings and semantic search (not just keyword matching)
- Elasticsearch's vector search capabilities (kNN search)
- Integration with ML pipelines for search personalization
- When to use vector search vs. traditional BM25 scoring
Beyond Search: The Analytics Engine
Elasticsearch is increasingly used for analytics, not just search:
- Time-series data aggregation for dashboards
- Real-time alerting based on log patterns
- Application Performance Monitoring (APM)
- Security Information and Event Management (SIEM)
Recruiter's Cheat Sheet: Spotting Great Candidates
Conversation Starters That Reveal Skill Level
Instead of asking "Do you know Elasticsearch?", try these:
| Question | Junior Answer | Senior Answer |
|---|---|---|
| "How would you improve search relevance for an e-commerce site?" | "Add more fields to search" | "Analyze query logs to understand user intent, implement synonym rings, use function scoring to boost by popularity and recency" |
| "Your cluster is running slow. Where do you start?" | "Add more nodes" | "Check slow query logs, analyze shard distribution, look for expensive aggregations or wildcard queries" |
| "How do you handle search for multiple languages?" | "Use a language analyzer" | "Depends on the use case—single index with per-field analyzers, language-specific indices, or a language detection layer" |
Resume Signals That Matter
✅ Look for:
- Specific scale metrics ("Indexed 500M documents," "Handled 5,000 QPS")
- Relevance work ("Improved search CTR by 30%")
- Production operations experience (upgrades, migrations, incident response)
- Mentions of Kibana, Logstash, or Beats (ELK stack familiarity)
- Experience with other search tech (Solr, Algolia) shows transferable skills
🚫 Be skeptical of:
- "Expert in Elasticsearch" without scale or complexity details
- Only development experience (never operated a production cluster)
- Listing every ELK component without depth in any
- No mention of monitoring, alerting, or performance tuning
GitHub Portfolio Signals
- Custom Elasticsearch plugins or analyzers (advanced)
- Search-related open source contributions
- Projects showing relevance tuning, not just basic CRUD
- Evidence of handling real data volumes (not just 100-document demos)
Common Hiring Mistakes
1. Requiring Elasticsearch for Simple Search Needs
If you're building product search for a 10,000-item catalog, you probably don't need Elasticsearch expertise. Algolia, Meilisearch, or even PostgreSQL full-text search might be simpler. Ask yourself: Do we need the complexity?
Better approach: Hire for search concepts. A developer who understands relevance, tokenization, and query optimization can learn any search engine.
2. Testing for Query DSL Memorization
Elasticsearch's Query DSL has dozens of query types with complex syntax. Nobody memorizes it all—they look up the documentation.
Better approach: Give a problem like "Users search for 'blue running shoes' but we want to also show navy jogging sneakers. How would you approach this?" Test thinking, not syntax.
3. Ignoring Distributed Systems Knowledge
Elasticsearch is a distributed system. Candidates who only know how to write queries will struggle with:
- Shard allocation and rebalancing
- Split-brain scenarios
- Consistency vs. availability tradeoffs
Better approach: Ask about experience with any distributed system—Kafka, Cassandra, or MongoDB at scale. The concepts transfer.
4. Overvaluing Cluster Administration
With managed services (Elastic Cloud, OpenSearch Service) handling infrastructure, pure ops skills are less critical than they were 5 years ago. Focus on search quality and application integration.
Better approach: Balance ops and dev. "Can you tune relevance?" matters more than "Can you configure JVM heap settings?" for most roles.
5. Not Testing Problem-Solving Under Ambiguity
Real Elasticsearch problems are ambiguous: "Search isn't returning good results" or "The cluster is slow sometimes." Strong candidates clarify requirements, ask probing questions, and gather data before proposing solutions.
Better approach: Present vague scenarios and see how they investigate. Do they ask about query patterns, data volume, user feedback?