Senior LLM Engineer – Agentic AI Platform

O projekcie: Project Description We are looking for a Senior LLM Engineer to contribute to the development of an advanced Agentic AI platform tailored for the construction domain. The platform focuses on deterministic, reliable, and production-ready LLM workflows supporting complex decision-making, structured reasoning, and domain-specific automation. The role involves designing robust architectures that ensure predictable outputs from large language models, implementing guardrails, and building evaluation pipelines. You will work on improving mathematical accuracy, enforcing structured outputs, and enabling scalable AI-driven services integrated with modern backend APIs and vector-based retrieval systems. Wymagania: - Strong experience as an LLM Engineer, Applied AI Engineer, or similar role - Proven expertise in designing deterministic LLM workflows and evaluation pipelines - Deep understanding of training dynamics, loss functions, and optimization strategies - Hands-on experience with parameter-efficient fine-tuning methods (LoRA, QLoRA) - Practical experience implementing structured outputs and validation using Pydantic - Strong knowledge of tool-calling mechanisms in agent-based AI architectures - Solid understanding of Model Context Protocol (MCP) architecture and implementation - Experience building production-grade LLM applications with FastAPI or Flask - Experience developing UI components using Streamlit - Strong understanding of REST principles, HTTP methods, WebSockets, and API lifecycle management - Experience working with relational and non-relational databases (PostgreSQL, MySQL, MongoDB, Redis) - Experience with vector databases such as Pinecone, Weaviate, or pgvector - Knowledge of semantic search and retrieval-augmented generation architectures - Strong problem-solving skills and ability to work in production-oriented AI environments Nice to Have - Experience in construction industry solutions or domain-specific AI applications - Experience with multi-agent orchestration frameworks - Knowledge of observability and monitoring for LLM s Codzienne zadania: - Design and implement deterministic workflows ensuring consistent and reliable LLM outputs - Develop evaluation frameworks to measure and validate deterministic model behavior - Address core LLM challenges, including improving mathematical reasoning accuracy and output reliability - Implement guardrails using structured outputs and Pydantic-based input/output validation - Design and manage tool-calling mechanisms for agent-based architectures - Build and optimize LLM pipelines using parameter-efficient fine-tuning techniques (LoRA, QLoRA) - Apply knowledge of training dynamics, loss functions, and optimization strategies - Implement and maintain Model Context Protocol (MCP) architecture - Develop and deploy LLM-powered services using FastAPI or Flask - Build lightweight front-end interfaces using Streamlit for AI-driven workflows - Design RESTful APIs, manage HTTP methods, and implement WebSocket-based communication - Oversee API lifecycle management, versioning, and performance optimization - Integrate structured and semantic data storage solutions (SQL, NoSQL, vector databases) - Implement semantic search and retrieval pipelines leveraging vector databases