Pujitha Sangaraju

About

Robotics software engineer focused on reliability in real-time sensor pipelines and ROS/ROS2 integration, with hands-on experience hardening systems against noisy data, dropouts, and network variability. Uses ROS introspection tools plus simulation (Gazebo/Webots) to diagnose latency and stability issues before hardware deployment, and supports repeatable rollouts via Docker and CI/CD.

Experience

Education

Key Strengths

• Designs reliable sensor-to-processing data flows for robotics systems
• Robust handling of sensor dropouts (timeouts, validity flags, last-known-good buffering)
• Improves data quality via filtering, thresholding, and outlier rejection
• Builds modular ROS/ROS2 nodes and integrates core ROS concepts (topics/services/params/launch)
• Diagnoses real-time latency issues using ROS tooling (rqt_graph, rostopic hz) and log analysis
• Optimizes real-time behavior by tuning message rates, callback behavior, and queue sizes
• Designs resilient distributed robot comms (heartbeats, stale-data rejection, safe local mode)
• Uses simulation (Gazebo/Webots) to surface timing/noise issues before hardware deployment
• Containerizes deployments with Docker and uses CI/CD to automate build/test/deploy
• Designed and maintained production FastAPI service with Postgres/Redis and centralized error handling
• Reliability engineering: strict timeouts, retries with exponential backoff limited to idempotent operations, idempotency keys/background jobs to prevent duplicate work
• AWS production delivery using API Gateway + Lambda with ECS Fargate for async processing
• Secure secrets/config management with Secrets Manager/SSM, KMS encryption, and least-privilege IAM per service/task
• Built AWS Glue ETL pipelines with schema evolution strategy, data quality checks, and failure alerting/triage via CloudWatch/SNS
• Legacy batch workflow modernization with parallel runs, checksums/metrics parity validation, incremental stakeholder sign-off, and phased rollout with rollback
• SQL performance tuning with measured impact (p95 latency ~2s to ~200ms via composite index + query rewrite)
• Designed and evolved a telemetry ingestion backend from monolith to decoupled event-driven pipeline
• Deep practical handling of at-least-once delivery with idempotency/deduplication to prevent duplicate writes
• Reliability engineering: DLQ handling, backoff, and failure-mode analysis during spikes
• API design for frontend enablement: contract-first schemas, versioned endpoints, consistent error formats, OpenAPI docs
• Production security implementation: JWT/OAuth, RBAC, tenant isolation, and Postgres RLS policies
• Led safe migrations with phased cutover (feature flags, shadow reads, canary traffic), dual writes, and integrity validation (checksums/counts)