Saaketh Chaganty

Built backend services in Java (Spring Boot) with PostgreSQL, Redis, Docker, AWS for an Agri-analytics platform used by 150+ clients. Implemented cloud-native solutions with containerization, infra-as-code, and automated deployment pipelines with rollback safety. Automated infra with modular Terraform stacks; reduced manual config ops, ensured reproducible deploys. Integrated weather APIs, soil IoT data, and satellite imagery to generate farm-health and yield predictions for 200+ operators. Implemented asynchronous GEE tile extraction & processing retry pipelines with REST APIs with workers processed 50,000+ tiles/day at >99% reliability. Added analytics + recommendation logic for crop insights; reduced agronomist manual review by 40%. Wrote JUnit + Mockito integration tests, improving reliability and reducing production bugs. Built function-calling and reasoning agents for remote-sensing interpretation; reduced agronomist analysis effort by 40%.

Built a real-time audio transcription pipeline using Whisper API, WebSocket streaming, and AWS, processing concurrent physician dictations from 50+ clinical sessions. Designed stateless services with caching, async queues, and cost-efficient architecture to scale with demand. Implemented a real-time doctor-patient co-pilot system with bidirectional WebSocket integration, enabling an intelligence engine with temporal audio segmentation, memory context, and structured FHIR extraction with 82%+ real-time event recall and sub-300ms alerting during sessions. Engineered a LangGraph-based agent mesh using ReAct and MCP tool calling with medical ontology constraints achieving 92% entity-F1 across Epic/Athena EHR feeds with DsPY prompt optimization. Built a Neo4j medical knowledge graph pipeline ingesting 100+ documents/day into ~50k nodes and ~200k relations, improving longitudinal pattern detection by 36%. Created server-sent events token streaming for LLMs at ~120 tokens/sec with adaptive th

Built event-driven trading infrastructure using AWS Lambda, API Gateway, DynamoDB, SQS, and Node.js/Python; supported more than 10,000 traders with under 120 ms latency and daily trading flows above $20,000. Developed REST and GraphQL services for trading, social feed, chat, and portfolio dashboards; delivered 99.9% uptime and reduced median latency by 35% through indexing and caching. Designed distributed feed system for crypto-social platform; scaled to over 50,000 concurrent users. Automated cloud deployment using Terraform, Docker, and CI/CD pipelines (GitLab and GitHub Actions); improved deployment frequency by three times and reduced AWS cost by 22%. Implemented idempotent trade execution, event deduplication, and transaction safeguards; lowered duplicate execution errors by more than 90%. Built WebSocket messaging for live market data and alerts with sub-second latency. Led a team of 8 engineers, reviewed architecture specifications, guided sprints, and enforced release standard