Shobhit Mehrotra

Monitoring and Alerting Team

Ancestry Knowledge Base MCP

• Create, extend and maintain data platform RESTful microservices and solutions for backend processing and provisioning of application performance metrics to monitor application health and generate alerts.
• Designed the data pipeline to gather real-time application performance metrics from NewRelic APIs for 1000+ Ancestry’s application stacks and saved the data to the AWS S3 data lake for further data processing.
• Built a GitOps-driven pipeline automating onboarding of enterprise knowledge into an internal Onyx/Senpai AI platform across a three-repo system (declarative YAML config, Python CLI engine, Terraform/EKS platform)
• Engineered S3-backed inter-stage state and registry for reference-safe reconciliation, drift detection, and idempotent delete flows, improving data integrity and operational reliability.
• Integrated cross-account IAM role assumption with LLM-backed AI enrichment system to auto-generate resource metadata (description, tags, category), enhancing resource management.

Oasis — Cloud Resource Inventory & Event Harvesting Platform

• Built and maintained an internal cloud CMDB that discovers, normalizes, and persists metadata for 80+ AWS and third-party resource types (EC2, ECS, EKS, RDS, Lambda, S3, IAM, CloudWatch, and more) into DocumentDB/MongoDB.
• Designed a schema-driven ingestion library (Pydantic + MongoEngine + JSON-QL mapping) covering the full resource lifecycle (create, update, delete, backfill), enabling new resource types to be onboarded via config instead of custom code.
• Developed the harvesting service on AWS Lambda and FastAPI with a multithreaded SQS poller (FIFO ordering, message grouping, retry queues) that consumes EventBridge/CloudTrail events, plus REST APIs for on-demand cross-account backfill and CloudWatch metric queries.
• Provisioned end-to-end infrastructure as code with Terraform (API Gateway, Lambda, SQS/DLQ, SNS, KMS, IAM, Secrets Manager) and containerized deployments with Docker, backed by CI/CD, automated testing (pytest/moto/mongomock), and semantic versioning.

Ancestry AI gateway and cost tracking platform:

• Engineered a centralized, multi-cloud AI Gateway (Python/LiteLLM) unifying customer-facing LLM access across AWS Bedrock, Azure OpenAI, Google Vertex/Gemini, and SageMaker behind a single OpenAI-compatible API.
• Implemented cross-provider fallback and routing with a modular per-provider config system, improving availability by rerouting failed requests to healthy deployments across regions without code changes.
• Deployed to AWS ECS/EKS via Terraform, Docker, and Harness CI/CD, with custom JSON metrics (latency, tokens, cost) streamed to New Relic and automated rollback for safe releases.

Cloud & AI Metrics Automation Platform:

• Developed sre-oasis-quests, a Python-based SRE automation platform running scheduled AWS Batch and Lambda jobs, exposed through a Typer CLI with a REST trigger API, to harvest usage, cost, and inventory data across cloud and AI infrastructure.
• Built metrics-harvesting pipelines that aggregate LLM gateway usage and cost data from multiple providers (Azure OpenAI, Gemini, LiteLLM, AWS Q) into PostgreSQL and MongoDB, enabling cross-provider spend visibility and cost-optimization reporting.
• Implemented automated audit and compliance jobs covering MSSQL database usage, load balancer certificate expiry, resource-tagging (untagged/mistagged/retired resources), and CloudWatch log-group activity, generating "quests" (actionable findings) from Oasis platform data to reduce manual review effort.

Other projects:

• Created, extended and maintained data platform RESTful microservices and solutions for backend processing and provisioning of application performance metrics to monitor application health and generate alerts.
• Designed the data pipeline to gather real-time application performance metrics from NewRelic APIs for 1000+ Ancestry’s application stacks and saved the data to the AWS S3 data lake for further data processing.

• Helped students with assignments, course projects, and exams.

• Patient Measurement Extractor & Dashboard Tool (Technology used: Java, MSSQL, MongoDb, MapReduce)

• HP Server License Tool (Technology used: Java, Java Swing, MSSQL, Shell Script)