Elan is a platform for building, deploying and scaling AI agents. A developer writes an agent, and the platform provides everything else needed to run it as a product. You would be one of the first engineers on the team.
As Knowledge & Data Engineer you own OCE, the context engine every agent on Elan asks for business meaning, and you bring each customer's meaning into it. OCE turns a company's documents and data sources into a governed vocabulary, and answers agents' questions with checked, sealed evidence. It is built and tested. Your job is to run it in production, connect it to the products launching from January, and onboard the design partners whose agents depend on it.
What you will own
Customer onboarding. Running induction over a customer's documents, reviewing what OCE proposes with their experts, and getting the vocabulary approved.
Metric definitions. Writing and versioning governed metrics with the customer, so agents answer from an agreed definition rather than a guess.
Live data. Mapping a customer's tables to the vocabulary, and running federated queries over their sources through Trino and BigQuery, behind a SQL firewall that refuses anything but a named, permitted table.
Knowledge graph. Neo4j with neosemantics, holding RDF, OWL and SKOS, with a second backend on Spanner.
Releases. Versioned ontologies and metrics, gated and staged releases, pinned agents and scheduled go-live, so a change to meaning never reaches an agent unreviewed.
OCE in production. Deploys through the runbook, per-tenant limits, audit stores that verify, and deleting a tenant's data when its account closes.
Connections to the platform. The token OCE reads, the usage the Ledger Service reads, and the tool rules the Action Policy Service holds.
What you will deliver in your first 90 days
By day 30. OCE running locally with its test suite green, a demo customer onboarded end to end in the Studio, and your first change shipped through the deploy runbook.
By day 60. The first design partner onboarded, and usage exported to the Ledger Service with totals that match.
By day 90. Action Policy rules enforced by OCE, an onboarding playbook others can follow, and you as primary owner of OCE in production.
How you will work
You report to the head of product and technology.
You work beside the Applied AI Engineer, who measures the answers your vocabularies and mappings produce.
You work with the Platform Engineer on the places the platform meets OCE, and with the DevOps Engineer, who owns the infrastructure OCE is deployed on. You join their on-call rotation.
You meet design partners' domain experts every week, with the Domain Ontology Advisor for their industry when one is engaged.
You are based in Noida with a hybrid working arrangement. Compensation is 30 to 70 LPA.
We hire for four things: ownership, curiosity, developer obsession and delivering results.
Minimum requirements
6 to 12 years in data engineering, knowledge engineering or backend engineering, with at least 3 building and running Python services or data pipelines in production.
Knowledge graphs or semantic layers in practice: RDF, OWL or SKOS and SPARQL, or a semantic layer such as dbt metrics, LookML or Cube.
SQL in depth across engines: PostgreSQL, a warehouse such as BigQuery, and ideally a federated query engine such as Trino or Presto.
Has mapped business terms to warehouse tables, and agreed definitions directly with business users.
PostgreSQL schema design and versioned migrations, with SQLAlchemy or a similar ORM.
Security in practice: tenant isolation, defences against SQL injection, and audit trails that can be verified.
Writes tests as part of the change, and has owned a service through production incidents.
Good to have
Neo4j, GraphDB or Stardog.
Domain experience in retail, hospitality or finance, and standards such as FIBO or GS1.
Data catalogues such as OpenMetadata, or virtual knowledge graphs such as Ontop.
Experience as an early engineer at a startup.
GCP.
