MLOps Engineer

MLOps Engineer (Remote) — AgileSoft (Outstaff) for UK FinTech Group

ABOUT AGILESOFT
- AgileSoft provides high-caliber engineers via outstaffing to global clients.
- We’re hiring an MLOps Engineer for a UK-based FinTech group (credit-risk models).
- Excellent English (spoken & written) is REQUIRED.

RESPONSIBILITIES
- Build & operate end-to-end ML platforms: ingestion → training → evaluation → deployment → monitoring
- Productionize credit-risk models (PD/LGD/EAD, scorecards, GBMs) with model registry, feature store, reproducible pipelines
- Implement CI/CD for ML (training & serving): champion–challenger, canary releases, rollbacks, automated tests
- Set up observability: latency/SLOs, model quality, drift (PSI), data-quality checks, alerting
- Ensure governance/compliance: versioning, lineage, approvals, reason codes, GDPR/PII handling
- Collaborate with Data Science, Risk, Data Engineering, Security; maintain docs & runbooks

MUST-HAVE SKILLS
- Excellent English; clear communication with UK stakeholders
- Production ML on Kubernetes + Docker; Python (FastAPI/Flask), Bash; Git & code reviews
- CI/CD for ML (GitHub Actions/GitLab CI/Jenkins), artifact management, environment promotion
- Model registry (MLflow) & feature store (Feast); experiment tracking & lineage (DVC/MLflow)
- Orchestration: Airflow (or similar) for training/inference pipelines & scheduled jobs
- Monitoring & logging: Prometheus/Grafana, ELK/OpenSearch; drift monitoring (PSI)
- Data quality & validation (Great Expectations or equivalent); schema control (pydantic/Marshmallow)
- Cloud (AWS/GCP/Azure), IaC (Terraform), secrets management (Vault/SM/SSM)
- Security & compliance for FinTech: PII handling, access control, audit logs, least-privilege

NICE-TO-HAVE
- Streaming/near-real-time scoring: Kafka/Kinesis; Faust/Spark Structured Streaming
- Kubeflow, Flyte, or managed ML platforms (Vertex/SageMaker/Azure ML)
- Evidently AI (or similar) for monitoring; Argo Rollouts; GitOps (Argo CD)
- Credit-risk/FinTech (IFRS 9, governance, reason codes)
- Performance optimization: A100/T4, mixed precision, Triton/ONNX/TorchServe
- Databases & warehousing: Postgres/ClickHouse/Snowflake/BigQuery; data lakes (Delta/Iceberg)

WHAT WE OFFER
- Remote-first, international team
- Competitive compensation, long-term engagement
- Opportunity to shape a robust ML platform for high-impact credit decisioning