Ermias Samson

Architected and shipped full stack applications using React/Next.js/Tailwind CSS on the frontend and FastAPI plus Node.js/NestJS microservices on the backend. Built LLM-powered backend services for document intelligence and conversational AI, including RAG pipelines with LangChain integrated with Pinecone and structured data sources. Developed multi-agent workflows using LangGraph for tool-calling and stateful orchestration, and used LangSmith for tracing, debugging, and evaluation to improve production reliability. Implemented async and streaming inference to reduce latency, with secure REST APIs (JWT/OAuth, rate limiting, retries, fallback strategies). Engineered ETL pipelines with PySpark/Databricks for preprocessing large datasets for LLM ingestion, deployed on AWS (EKS/Lambda/S3), and improved uptime to 99.8% using monitoring and observability (CloudWatch/logging/tracing). Built CI/CD automation with Kubernetes/GCP to reduce deployment time by 65%.

Built scalable backend systems and RESTful APIs for AI-driven, data-intensive applications using Python (Django/Flask/FastAPI) and Node.js/NestJS. Led frontend development across React/Next.js/Vue/Angular, optimizing performance to sub-200ms load times and improving consistency with Tailwind CSS, SASS, and Material UI. Implemented early-stage LLM integrations with LangChain, creating modular prompt pipelines and API-based inference workflows. Developed PySpark-based ETL pipelines for batch processing and large-scale transformations, and deployed containerized services (Docker/Kubernetes/EKS) with CI/CD. Applied TDD with Pytest/Jest, implemented observability for distributed systems, and improved query response times by 20% by adding GraphQL APIs. Optimized AWS infrastructure (VPC, load balancing, SSL, auto-scaling) and reduced deployment time by 40% using AWS EKS with CodeBuild and CodePipeline. Improved app speed by 40% through code-splitting and Redis caching.