Elias Müller

Internal RAG/LLM and backend systems in Swiss retail banking, over 1,000 daily users.

Development of Python RAG pipelines with hybrid keyword/vector search over indexed bank documents for business-assisting chat workflows. Achieving 98% correctness in responses on an evaluation set of around 200 conversations validated by subject matter experts, through hybrid retrieval tuning, response validation workflows, and iteration on persistent issues. Development of LangGraph workflows for retrieval routing and multi-stage searches over Elasticsearch indexes from multiple extensive document sources. Implementation of multi-model agent orchestration and custom MCP servers for tool access, with A2A interfaces for agent-to-agent and human-in-the-loop interactions. Development and maintenance of the entire stack for internal chat applications, including Angular UIs, Kotlin/Python services, OpenShift operation, and low-latency streaming via WebSocket/SSE

Tech: Python, Kotlin, Spring Boot, Kafka, LangChain, LangGraph, A2A, MCP, FastAPI, Pydantic, Azure OpenAI, Azure AI Search, Elasticsearch, LLMOps, asyncio, Angular, OAuth2, JWT, MySQL, Redis, Prometheus, Grafana, Splunk, OpenShift, Argo CD

Event-driven full-stack application for Swiss Customs for import/export tracking of goods.

Development of web applications and RESTful microservices in Angular and Java for managing flagging workflows and customs simulations. Extension of the Drools rule engine for detecting complex relationships between goods metadata and other customs-relevant datasets. Reactive backend for processing millions of Kafka events per week

Tech: Java, Angular, Drools, Kafka, Spring Reactor, Docker, Spring Boot, PostgreSQL, Bash

Backend infrastructure for managing firmware installation of traffic monitoring devices in connected transport systems.

Development of new and maintenance of complex existing backend microservices (14 in total) for managing firmware installation of traffic monitoring devices, with integration of internal UI frameworks and multiple data sources via REST and SOAP; operation in Kubernetes on AWS. Connection of about a dozen traffic monitoring devices (test stage) to the platform via MQTT. Implementation of geometric algorithms for vehicle positioning based on telemetry between road devices and in-vehicle units. Creation of Kubernetes configurations and Helm charts for the microservices. Improvement of device authentication and service security, including SSL certificate automation and fixing security vulnerabilities.

Tech: Java, Kotlin, Python, Spring Boot, GitLab, Kubernetes, Helm, AWS, MQTT, ActiveMQ, Selenium, JUnit, Mockito, REST, SOAP