Senior LLM / AI Engineer

Join the core engineering team building a next-generation enterprise AI platform designed to transform massive volumes of unstructured domain data into structured, actionable intelligence. As a Senior LLM / AI Engineer, you will bridge the gap between bleeding-edge AI research and hardened software engineering. You will architect the platform's core AI orchestration layer from scratch—engineering multi-agent workflows, robust Retrieval-Augmented Generation (RAG) pipelines, and deterministic reasoning guardrails inside a production-grade, asynchronous FastAPI backend.

Details

Role: Senior LLM / AI Engineer

Seniority: Senior (4+ years software engineering / 2+ years production LLM experience)

Allocation: Full-time (1 FTE)

Operating Model: Enterprise Platform Engineering

• Responsibilities
• Agentic AI & Multi-Agent Orchestration
• Design and build collaborative multi-agent systems utilizing LangGraph, LangChain, or custom state machines featuring conditional routing, advanced memory management, and human-in-the-loop validation steps.

Build secure tool-calling infrastructure complete with runtime registries, execution timeouts, error recovery loops, and standardized Model Context Protocol (MCP) server integrations.

Advanced RAG & Knowledge Retrieval

Construct scalable ingestion pipelines implementing advanced chunking strategies, hybrid search (semantic vector + BM25 keyword), and multi-hop document reasoning.

Optimize retrieval fidelity using sophisticated re-ranking models, query decomposition, and strict citation grounding algorithms to completely mitigate hallucinations.

Deterministic Workflows & Output Validation

Force non-deterministic LLM engines to yield 100% reliable outputs using strict Pydantic validation, JSON Schema constraints, and structured function calling.

Code hybrid runtime pipelines that elegantly decouple deterministic business logic from LLM-powered abstract reasoning.

Evaluation, Testing & Observability

Deploy automated evaluation frameworks and continuous regression benchmarks to programmatically grade output accuracy, context relevance, and token consumption.

Implement production-level observability (e.g., via Langfuse or LangSmith) to trace multi-step agent decisions, track token budgets, and manage caching layers to balance cost and latency.

Requirements

Required Qualifications

Experience Baseline: 4+ years of core software engineering, with at least 2 years of hands-on experience shipping LLM-powered applications directly to production (beyond prototypes or notebooks).

Python Engineering: Strong proficiency in Python, writing clean, asynchronous, production-ready code natively using FastAPI.

Orchestration Frameworks: Deep practical knowledge of building stateful, multi-turn graphs with LangGraph or custom graph-based state networks.

The Production Stack:

Databases: Advanced knowledge of PostgreSQL and Redis for state tracking and caching.

Vector Infrastructure: Hands-on experience with at least one enterprise vector database engine (pgvector, Qdrant, FAISS, Pinecone, or Weaviate).

Validation Rigor: Expert use of Pydantic for type systems and schema constraints.

Preferred Experience

Production experience architecturalizing Model Context Protocol (MCP) systems.

Familiarity with hybrid search integration using Elasticsearch.

Background architecting multi-tenant AI products utilizing organization-level data isolation, token rate-limiting, and tenant-based cost attribution.

Pragmatic understanding of parameter-efficient fine-tuning (LoRA, QLoRA) and tracing frameworks.