Siddharth Yadav

Implemented an LLM agent system on LangGraph to triage security alerts and auto-generate detection rules, reducing mean time to detect by 50%. Built reusable RAG ingestion and retrieval pipelines on LlamaIndex and ChromaDB supporting vector and knowledge-graph modes, citation-aware querying, and incremental hash-based re-indexing. Designed a tool-first agent platform enabling composition of specialist AI assistants from a shared tool registry (RAG, SIEM, formal verification, web search, fine-tuning), with a provider-abstracted LLM layer and MCP server exposure for AI-IDE integration.

Architected an agentic workflow on the ACS Publications integration platform using LangGraph's ReAct loop, orchestrating multi-step API integrations across AWS services and third-party SaaS vendors, reducing edge-case integration failure resolution time by 70%. Engineered an agentic GitHub PR analytics system to eliminate manual identification of high-risk PRs by automating data collection, cutting review effort by 85%. Set up a GitHub Actions CI/CD pipeline for ACS Pubs repos, increasing deployment frequency from weekly to over 10 deployments per day.

Developed backend services of a highly scalable portal for creating and modifying policies using Lambda, RDS, NodeJS, S3, achieving a 90% improvement in TAT. Built a cross-organization report and access management platform using ReactJS, EC2 and ExpressJS, reducing report generation time by 80%. Integrated the legacy Genesis Configurator API into the backend system, and implemented product APIs with authorization and rate limiting using AWS API Gateway, VPC, Redis — enabling new vendor integrations and partnerships.

Built a data sync pipeline for a banking client to sync customer and loan records across RDS, S3, and downstream systems, implementing incremental batch processing with offset-based pagination and automated status tracking. Designed and implemented a retrying microservice for syncing cases to downstream leveraging Step Functions and Redis, reducing failure rate by 70%. Mentored new employees on clean code and best practices.