Prakash Nidhi Verma

About

Backend/platform engineer with experience across edtech, logistics, and AWS internal systems—owned a production course recommender end-to-end (model serving + APIs + caching/observability), delivering +30% CTR and -20% latency. Has scaled real-time delivery visibility/rerouting on Kubernetes/EKS to sub-200ms P95 during demand spikes and built billion-events/day telemetry pipelines on AWS (Kinesis Firehose, Lambda, S3, Redshift) with schema evolution, dedupe, and replay support.

Experience

Education

Key Strengths

• Owned end-to-end production backend feature (course recommender) from scoping through on-call operations
• Staged rollout using feature flags with KPI monitoring (CTR, session duration) before full launch
• Built strong observability (CloudWatch/Grafana) tracking API latency, cache hit ratio, and model inference time
• Resolved production incident caused by Redis key expiration misconfiguration; improved resilience with circuit breaker/retries
• Improved recommender outcomes: +30% CTR and -20% average latency via model refresh and caching strategy tuning
• Scaled real-time logistics platform with reliability patterns (circuit breaker, exponential backoff) to prevent cascading failures
• Performance optimization: reduced P95 latency by 24% via caching and DB/query refactors; maintained sub-200ms P95 at peak
• Designed high-volume telemetry pipeline (Kinesis Firehose + Lambda + S3/Redshift) with schema versioning, dedupe, and replay/backfill support
• Used metrics/logging to quickly diagnose DynamoDB throttling and hot partitions; mitigated with access-pattern changes and backoff