Staff Machine Learning Engineer | Agentic Systems, MLOps and Retrieval

Tekion is building AI into dealership software, and this is the senior engineering seat behind it. The requirement list is one of the most complete MLOps specifications in today's edition and it describes a genuinely modern production ML stack rather than research: LLMs, retrieval systems, vector stores and graph or knowledge stores, orchestration through LangChain, LlamaIndex, function calling or AgentKit, and agent architectures including reactive, planning and retrieval augmented agents with safe execution patterns. Around that sits the operational half most postings skip, namely experiment tracking, model registries such as MLflow, feature stores, A/B and shadow testing, drift detection, and observability through OpenTelemetry, Prometheus and Grafana. Pipelines run on Airflow or Kubeflow with Spark or Flink and Kafka or Kinesis. At 8 to 11+ years this is a deep seat, and cost optimisation on AWS is named explicitly.

Tekion asks for 8 to 11+ years in ML engineering and MLOps, or backend and platform engineering with production ML. The published qualification list is long and specific:

Experience with LLMs, retrieval systems, vector stores and graph or knowledge stores. Strong software engineering fundamentals: Python plus one of Java, Go or Scala, API design, concurrency and testing. Hands on with orchestration frameworks and libraries such as LangChain, LlamaIndex, OpenAI function calling and AgentKit. Knowledge of agent architectures (reactive, planning, retrieval augmented) and safe execution patterns. Pipelines and data: Airflow or Kubeflow or similar, Spark or Flink, Kafka or Kinesis, with strong data quality practices. Microservices and runtime: Docker and Kubernetes, service meshes, REST and gRPC, plus performance and reliability engineering. Model ops: experiment tracking, registries such as MLflow, feature stores, A/B and shadow testing, and drift detection. Observability: OpenTelemetry, Prometheus and Grafana, including debugging latency, tail behaviour and memory or CPU hotspots. Cloud: AWS preferred (IAM, ECS and EKS, S3, RDS and DynamoDB, Step Functions and Lambda) with cost optimisation experience. Security and compliance: secrets management, RBAC and ABAC, and PII handling.

Location: Bangalore HQ.

Honest fit guidance: this is an engineering role with ML in it, not a modelling role. Nothing in the requirement list asks you to design novel architectures or publish; almost all of it is about running models reliably and affordably in production. Data scientists who want to move into engineering will find the operational half unfamiliar. Platform and backend engineers who have shipped ML systems are, by Tekion's own wording, explicitly welcome, since it accepts backend or platform engineering with production ML as an alternative to an ML engineering title.