AI / ML Platform Engineer

• Build and maintain a large‑scale knowledge graph (entities, transactions, ownership, sanctions, registries). Design ETL pipelines, entity‑resolution logic, and network‑level feature jobs for GNN models that score shell companies, detect circular flows, assess sanctions proximity and community risk.
• Create end‑to‑end pipelines that convert regulatory/investigative text into vector embeddings, store them in a vector DB, and orchestrate LLM‑based narrative generation, multi‑agent reasoning, and on‑demand query answering. Manage LLM gateways, prompt engineering, cost tracking and fallback handling for uncertain alerts.
• Own the non‑ML micro‑services that make ML outputs actionable (Alert Service, Case Manager, API Gateway, tuning matrix, deduplication, Signal Aggregator, rule‑engine integration, feature‑factory APIs, model registry). Ensure high‑throughput, low‑latency, event‑driven processing with Kafka, PostgreSQL, Redis and Docker deployments.
• Apply rigorous software‑engineering practices (CI/CD, MLflow, automated testing, monitoring, model governance, audit logging) to keep models robust, calibrated, explainable and compliant.
• Manage the on‑prem Kubernetes ecosystem (multi‑namespace clusters, Helm, ArgoCD GitOps, resource quotas, GPU scheduling) and the full data‑processing stack: Apache Spark/PySpark (batch & streaming), Flink/Kafka Streams, dbt incremental models, Terraform/Pulumi IaC, and observability (Prometheus‑Grafana, ELK). Handle node provisioning, container images (Docker, ECR/Harbor) and end‑to‑end logging/monitoring/alerting.
• Collaborate with distributed data‑science, compliance, investigation and platform teams to translate domain expertise into scalable, production‑ready AML solutions that continuously improve BNPP’s risk‑management posture.

Technical & Behavioral Competencies

• Advanced Machine‑Learning Engineering: Design, train, and ship production‑grade models (GNNs, LLMs, RAG pipelines) that are well‑calibrated, explainable and meet strict AML governance requirements. Demonstrates a deep understanding of model lifecycle management, performance monitoring, and continuous improvement.
• Graph‑Data Mastery: Expertise in knowledge‑graph modelling (Neo4j/Cypher or similar), graph‑ETL at scale, and graph‑neural‑network frameworks (PyG, DGL). Able to engineer high‑throughput network‑feature pipelines, optimize sub‑graph extraction (< 200 ms), and apply graph algorithms for risk scoring.
• LLM & Retrieval‑Augmented‑Generation Proficiency: Hands‑on experience with LangChain/LangGraph, prompt engineering, multi‑agent design, and vector‑database ecosystems (Pinecone, pgvector, Weaviate). Capable of building robust RAG architectures, chunking, embedding, retrieval, re‑ranking and integrating LLM gateways with cost‑control and fallback mechanisms.
• Backend & Integration Engineering: Solid track record building high‑availability micro‑services (FastAPI/Node.js), event‑driven pipelines (Kafka, Avro, Schema Registry), and rule‑engine integrations (Drools or custom Python). Skilled at designing state‑machines, de‑duplication engines, and API‑gateway (Kong) configurations that expose ML outputs to downstream investigators.
• Cloud‑Native & Data‑Platform Ops: Deep familiarity with on‑prem Kubernetes (multi‑namespace, Helm, ArgoCD GitOps), CI/CD pipelines (Jenkins, GitLab CI, GitHub Actions), and observability stacks (Prometheus‑Grafana, ELK/Loki). Proficient in Spark/Delta Lake, Flink/Kafka Streams, Terraform/Pulumi, and Docker containerization, including GPU node provisioning and sandbox quota management.
• Data‑Engineering Fundamentals & Governance: Adept at handling noisy, large‑volume data: incremental ETL, schema evolution, ACID compliance, and feature‑store patterns (Feast, Redis). Champions data‑quality, lineage, auditability, and regulatory compliance throughout the pipeline.
• Collaboration & Communication
  Excels in globally distributed, cross‑functional teams; translates complex ML concepts into actionable requirements for compliance officers, investigators, and business stakeholders. Communicates clearly in written and oral forms, writes comprehensive documentation (OpenAPI/Swagger, design specs), and mentors junior engineers.
• Problem‑Solving Mindset & Adaptability
  Approaches ambiguous AML typologies with curiosity and a systematic, data‑driven methodology. Quickly learns new financial‑crime domains, navigates regulated environments, and iterates on solutions under tight timelines while maintaining high quality and security standards.