Rauf's CV
Rauf Aghayev
Senior Platform & Backend Engineer
Profile summary
Senior platform and backend engineer with 10+ years at AWS, Miro and Fin AI, building the infrastructure behind production LLM and cloud platforms serving 100M+ users. Owns LLM routing and multi-provider failover (Bedrock, Azure OpenAI), embedding serving on SageMaker, and multi-region AWS platforms end to end, from architecture to on-call. Delivered $120K+ in annual cloud savings, cut incident response times by 30%, and launched services across EU, US and Australia regions with zero data loss. Conducted 50+ technical interviews at AWS.
Career highlights
Data Store Migration: Re-platformed the primary user data store from Redis to PostgreSQL, cutting over 165M records with zero downtime.
Infrastructure Cost Savings: Delivered $120K+ in annual savings through infrastructure cost investigations and resource reallocation at Fin AI.
System Modernization: Migrated legacy monoliths to Kubernetes, reducing deployment times by 40% and tuning autoscaling, rolling releases, and resource limits.
Technical Certifications: Earned the Certified Kubernetes Application Developer credential in 2022.
Key skills
Professional experience
Directs core infrastructure and machine learning deployment functions, managing model routing architectures and platform permissions. Oversees engineering operations focused on high-availability routing, cost reduction, and security controls across internal services.
- Delivered 120K dollars in annual savings through infrastructure cost investigations and resource reallocation.
- Cut memory by 1.17 GB per worker by deduplicating spaCy pipelines, worth roughly 50K a year.
- Onboarded new LLM models and hardened the routing layer, adding pricing integration, Azure failover, and removing single points of failure.
- Migrated Stella, E5, custom-answer and snapshot embeddings to SageMaker TEI behind feature flags, refactoring endpoint resolution to eliminate local model loads.
- Built the group-based permissions platform for Internal Tools, including access redesign, audit logging and authorization narrowing.
Directed backend service operations across global regions and AWS infrastructure, establishing system observability practices and compliance implementations for enterprise customers.
- Re-platformed the primary user data store from Redis to PostgreSQL, cutting over 165M records with zero downtime.
- Relocated EU datasets into the US region under compliance constraints with zero data loss, and brought backend services live in AWS Australia to cut latency for APAC customers.
- Built the AWS Cost Calculator integration, letting enterprise customers estimate AWS spend directly inside Miro.
Directed software engineering functions for enterprise cloud systems, managing technical delivery and engineering standards across distributed teams.
- Engineered front-end and back-end components for the AWS billing platform, improving billing accuracy.
- Sustained 99.9% uptime on the on-call rotation through monitoring and troubleshooting.
- Conducted 50 technical interviews, contributing to hiring-bar calibration and global team growth.
Managed core mail services processing high-volume traffic for consumer platforms with tens of millions of users, participating in on-call rotations and developing anti-spam logic.
- Migrated legacy monoliths to Kubernetes, reducing deployment times by 40% and tuning autoscaling, rolling releases, and resource limits.
- Designed and maintained 15+ fault-tolerant microservices for gmx.de, web.de, and mail.com while building Kafka-based event pipelines.
- Built a full-stack known-sender database that improved email deliverability, alongside user feedback features that lifted engagement by 15%.
- Hardened authentication mechanisms and secrets handling protocols while executing dependency vulnerability remediation across services.
- Built and maintained CI/CD pipelines for AWS deployments, instrumenting services with metrics, dashboards, and alerting.
- Decomposed monolithic infrastructure into microservices, increasing system scalability by 50%.
- Added automated integration test coverage as part of the microservice split.
Education
Computer Science | Focus: Machine Learning and Semantic Web. Thesis: Spreading-Activation-Based Search in RDF Data (supervisors: Prof. Sören Auer, Dr. Axel-Cyrille Ngonga Ngomo).
- Built a keyword entity-search engine over DBpedia in Java: 119 GB of RDF loaded into a 296 GB Neo4j reified graph with a Lucene dictionary, using Apache Jena and Stanford NLP - no SPARQL generation.
- Cut average query time from ~10 minutes to 27 seconds (40x) through Jaro-Winkler candidate pruning, with identical answer sets; evaluated on QALD-5.
Computer Science
Auer, Dr. Axel-Cyrille Ngonga Ngomo).
- Built a keyword entity-search engine over DBpedia in Java: 119 GB of RDF loaded into a 296 GB Neo4j reified graph with a Lucene dictionary, using Apache Jena and Stanford NLP — no SPARQL generation.
- Cut average query time from ~10 minutes to 27 seconds (40x) through Jaro-Winkler candidate pruning, with identical answer sets; evaluated on QALD-5.
Certifications
Languages
Azerbaijani · Native
Turkish · Native
English · C1
German · B2