Manaswini Gogineni

About

Backend engineer at Electric Hydrogen who built a serverless device-log ingestion and processing platform in Python/Flask, scaling throughput (4x peak ingestion) while keeping sub-300ms API latency. Strong in Postgres/SQLAlchemy performance (partitioning, materialized views) and production ML integration (ONNX model served via FastAPI microservice with async batch inference, Redis feature caching, and drift monitoring via S3/Lambda). Experienced designing secure multi-tenant systems with schema-per-tenant isolation and KMS-backed encryption.

Key Strengths

• Designed maintainable Flask APIs using blueprints + service-layer separation (improved testability/onboarding)
• Built high-throughput async ingestion pipeline with SQS + retry/DLQ; achieved 4x peak ingestion while maintaining sub-300ms API responses
• PostgreSQL/SQLAlchemy performance tuning (indexes, partitioning, bulk inserts, N+1 reduction) scaling to millions of log entries
• Used materialized views to cut analytics load times from seconds to under 200ms
• Production ML integration via separate FastAPI inference microservice + ONNX, with async batch inference through queues
• Reduced ML latency from ~500ms to ~120ms using Redis feature caching
• Implemented multi-tenant isolation with schema-per-tenant, tenant-aware middleware, per-tenant quotas, and KMS-based encryption controls
• Optimized background consumers with batching (25–30 msgs), Redis metadata caching, concurrency tuning, and backpressure to protect Postgres
• Led rollout of an opinionated-but-flexible internal CLI to standardize AWS infrastructure provisioning across multiple teams/orgs
• Designed programmatic generation of Terraform modules from validated architecture templates (VPC/EKS/RDS/IAM/networking) with secure defaults
• Drives adoption through pilot partnerships, strong documentation/templates, and low-friction CI/CD integration
• Systematic distributed-systems debugging using scoping, correlation IDs, and observability to pinpoint ingestion pipeline failures
• Implemented DLQ-based capture of malformed/edge-case messages to prevent silent data loss and harden pipelines
• Bridges on-prem industrial data sources with cloud/Kubernetes; builds real-time IoT telemetry pipelines feeding analytics

Reference Highlights

• Efficient execution
• Proactive involvement
• Ensures outputs align with goals
• Professional-level technical depth (Python/Flask/SQLAlchemy)
• Built and deployed a Flask API server integrated with PostgreSQL
• Reduced query latency via efficient queries
• Designed scalable, maintainable architecture still in use
• Takes ownership when others lack context
• Self-directed learning (extensively read documents)
• Improves robustness through unit tests
• Strong cross-functional communication to distinguish meaningful metrics from noise
• Improved data quality and downstream prediction accuracy
• Handles ambiguity well
• Delivers under tight deadlines (beta model within a week)

Experience

Education

Awards

• Great Workmanship award - Cisco ThousandEyes
• Quarter Recognition - Cisco ThousandEyes

Skills

Languages

Certifications