Devashish Taneja

Own device onboarding and registration services for Alexa Built-in (3P ABI) — the activation path for millions of third-party Alexa devices (TVs, soundbars, speakers) across multiple device manufacturers. Driving internationalization of Alexa+ for third-party devices across 9 locales (~2MM monthly active accounts): authored the technical design and am coordinating cross-team delivery across multiple service teams (capability publishing, locale management, globalization). Leading design and rollout of local-network device discovery via mDNS/Bonjour, replacing manual setup steps and reducing pairing failures during third-party device onboarding by 80% Designed and shipped Alexa+ opt-in/consent flows for the Alexa TV app, reconciling privacy and regulatory constraints with staged, experiment-gated rollout to production Serve in the team's on-call rotation for device-registration services: production incident response, root-cause analysis, and remediation.

Built a horizontally scalable Site Search service on OpenSearch with Kafka-based real-time synchronization from multiple upstream services — cut P95 full-text search latency from 30s to <50ms (600x); adopted as the org-wide search layer Designed a Fleet Upgrade orchestration system executing remote bulk upgrades across 100,000+ customer sites/day with idempotent retry semantics, fault-tolerant execution, and zero-downtime guarantees Led decomposition of a monolithic Ruby on Rails application into modular service boundaries and migrated AMI-based deployments to Docker — cut release cycles from 2+ months to <2 weeks, unblocking parallel delivery across teams Developed a Spring Cloud Gateway adopted by 20+ microservices — canary deployments, user-based routing, and role-based access control; became the standard ingress for the org Built an A/B experimentation platform (Spring Boot, MySQL, Redis) powering feature flags and alpha/beta cohorts at 5ms P99, enabling safe incremental rollouts across product teams Implemented a no-code rule engine evaluating condition-action rules at 10,000 executions/sec inside FaaS workflows (warranty, event processing) — removed engineering from the loop for business rule changes Built observability and support tooling (Rails, MongoDB) monitoring 2M+ energy sites at 1,000+ req/sec for faster issue triage, and cut CI feedback 8x (2h → 15min) by parallelizing test execution across Docker containers Implemented Docker container labeling with parent Job ID for periodic/dispatch jobs, improving workload traceability for operators of large clusters Added configurable job version retention to bound cluster state growth in production deployments Contributed Spring annotation-based background job scheduling (@Job), simplifying async task processing for Spring Boot applications