AI Systems Engineer Full Time

We aren’t building a chatbot; we are building a cognitive system that understands the complex web of global procurement. Your role is to bridge the gap between raw data and high-stakes financial decisions. You will design the agentic workflows that parse invoices, the Knowledge Graphs that map supplier relationships, and the rigorous evaluation frameworks that ensure our risk scores are mathematically sound.

The Responsibilities

·     Multi-Agent Workflows: Design and deploy autonomous agentic systems using LangGraph, PydanticAI, or CrewAI to handle complex, multi-step procurement tasks

·     API Integration & MCP: Use modern protocols like Model Context Protocol (MCP) to securely connect LLMs to internal tools and databases.

·     Modern RAG: Maintain a high-performance RAG pipeline that combines traditional vector search with semantic retrieval.

·     Graph-RAG Architecture: Transform unstructured data from PDFs and images into a structured Knowledge Graph (using Neo4j or AWS Neptune) to map entities, subsidiaries, and risk contagion.

·     Anomaly Detection: Design logic that goes beyond simple threshold alerts to identify “soft” risk signals, such as subtle shifts in supplier billing behavior or hidden corporate linkages.

·     The Scientist Mindset: Own the quantitative evaluation of our AI. You will build the test suites to measure Precision, Recall, and F1-scores for every model and agent we deploy.

·     LLM-as-a-Judge: Implement automated evaluation frameworks (using tools like Ragas or LangSmith) to grade agent reasoning chains for faithfulness and relevancy.

·     Targeted Fine-Tuning: While secondary, you will perform LoRA/QLoRA fine-tuning on open-source models (like Llama 4 or Mistral) to optimize specialized document extraction where off-the-shelf APIs fall short.

The Technical Stack

·     Languages: Expert Python (async, Pydantic, FastAPI).

·     AI Frameworks: LangGraph, LlamaIndex, and Hugging Face Transformers.

·     Graph Systems: Neo4j (Cypher) or AWS Neptune.

·     Evaluation: Arize Phoenix, Giskard, or Weights & Biases.

·     Infrastructure: Experience working alongside software teams in Docker/K8s environments.

How You Fit into the Team

You are the “Prefrontal Cortex” of our product.

·     Our Data Engineers provide clean data streams.

·     Our Software Engineers build the UI and deterministic business rules.

·     You build the reasoning layer that makes sense of the data and provides the “Explainable AI” insights that our users rely on.

Ideal Candidate Profile

·     Systems First: You think in loops, states, and graphs, not just prompts.

·     Data-Obsessed: You would rather spend a day building a better benchmark than an hour tweaking a UI.

·     Product-Minded: You understand that an AI model is only as good as the business decision it enables.

·     Experience: 5+ years in engineering, with a significant portfolio of shipped AI/Agentic systems in high-stakes domains (FinTech, LegalTech, or Supply Chain).