Sowmya Pandey

- Built and integrated GenAI-assisted workflows using Claude and other AI tooling to improve customer support experiences and internal engineering productivity across scalable platform systems. - Owned external account GraphQL backend using Java and Kotlin, supporting 1.3M+ accounts and 5M+ daily operations with less than 0.7% failure rate. - Developed account linking, retrieval, and deletion APIs with React-based UI flows, improving successful completion rates by 22% across member journeys. - Designed and optimized GraphQL schemas and routing identifiers, reducing malformed consumer requests by ~35% across 14+ dependent microservices. - Led on-call rotations for business-critical services, triaging 50+ high-severity incidents weekly and driving durable fixes through root cause analysis. - Built observability and diagnostics using Datadog metrics, logs, and alerting dashboards, reducing mean time to resolution from ~2 hours to 18 minutes. - Stabilized Kraken development environments, reducing integration blockers by 25% and improving QA and release validation throughput by ~30%. - Mentored interns and junior engineers, contributing to design discussions and improving overall engineering execution quality.

- Engineered low latency entitlement microservices using Java 17 and Spring Boot, reducing authorization decision latency by approximately 30 percent at peak concurrency. - Built React based entitlement and access control interfaces, improving approval resolution time by approximately 25 percent. - Orchestrated entitlement data ingestion and audit workflows using Apache Airflow DAGs, ensuring reliable and repeatable reporting across regulated domains. - Designed REST APIs with full audit traceability supporting approximately 25 million entitlement evaluations per month. - Integrated Apache Druid analytics to enable sub second aggregations for audit and compliance reporting. - Containerized services using Docker, Kubernetes, and Helm, reducing environment drift by approximately 60 percent.

- Designed and operated Kafka-based distributed services supporting advertising campaign billing, reconciliation, and revenue workflows across 200M+ monthly events. - Designed low latency aggregation APIs unifying legacy mainframe, Oracle, and SQL data sources, reducing lookup latency by approximately 20 percent. - Deployed containerized services on AWS ECS and Fargate to improve scaling behavior during peak ingestion periods. - Built Elasticsearch and Kibana dashboards enabling earlier anomaly detection and faster root cause analysis.

- Owned on device C++ audio processing components for Alexa, optimizing DSP pipelines and memory usage to improve responsiveness on constrained hardware. - Reduced latency and CPU utilization through concurrency safe scheduling and optimized buffer management across millions of devices. - Improved wake word detection accuracy by approximately 40 percent through model pruning and quantization. - Automated multi region firmware deployments using GitHub Actions and CloudFormation, reducing deployment overhead by approximately 30 percent. - Collaborated with ML, firmware, and backend teams to align device side optimizations with system wide telemetry.

- Built Python-based calibration and image-alignment pipelines using OpenCV, NumPy, SciPy, improving laparoscopic tool-tracking precision by 20%. - Deployed real-time ECC + motion-compensation algorithms on GCP (Compute Engine + Cloud Functions) to support scalable, HIPAA-compliant surgical imaging workflows.