This is a hands-on building role: you turn raw, messy fabrication data into the clean, well-modeled, AI-ready datasets that our AI/ML and analytics workloads run on
Responsibilities:
Build and operate ingestion, ELT/ETL, and orchestration pipelines that move data from our MongoDB Atlas operational store and other sources into our analytical and AI-serving layers
Implement layered (medallion-style) transformations with idempotent, backfillable, incrementally loaded jobs
Apply deduplication, normalization, and validation so downstream data is high-quality and trustworthy
Modernize legacy / homegrown data flows via incremental, strangler-fig migrations that keep production stable
Build embeddings and vector pipelines, and the feature/retrieval-ready datasets that RAG, semantic search, and agentic workloads depend on
Make production data AI-ready in practice: well-structured, lineage-tracked, and retrieval-friendly, in partnership with ML and application engineering
Implement real-time and change-data-capture flows from MongoDB (Change Streams / CDC) where workloads require fresh data
Implement the canonical data model, schemas, and data contracts defined by the Data Architect — enforced in-repo so other teams build against stable definitions
Exercise sound persistence judgment in execution: land data in the right store (document / NoSQL, vector, analytical) per the architectural direction
Contribute to build-vs-buy decisions by prototyping with proven, industry-standard tooling over custom development
Establish testing, data-quality, and lineage checks for the pipelines you own, with clear alerting and runbooks
Instrument pipeline observability (freshness, volume, schema-drift, cost)



so failures are caught before consumers feel them
Use AI-assisted development tools (Claude Code, Copilot, Cursor) as a force multiplier for transformation logic, query tuning, and migration scripting
Partner with database engineering on extracting from and protecting the production store
Partner with the Data Architect on implementing target-state patterns and surfacing what's hard to build
Partner with ML, AI, and application engineers on the data they consume — shaping and governing it so it's safe and ready to build on
If you have:
5+ years of hands-on data engineering experience building and operating production data pipelines at scale
Strong programming and data skills: Python and SQL, with solid software-engineering fundamentals (version control, testing, CI) — shipping and maintaining production code, not just notebooks
Hands-on MongoDB at production scale (Atlas idóneo): document modeling, aggregation framework, change streams / CDC, and extracting from a document store into analytical / AI-serving layers. Our stack is NoSQL / MongoDB, not relationa
l
, this is a core requirement, not an extra
Demonstrated experience with ELT/ETL pipeline design, transformation frameworks (dbt or equivalent), and orchestration (Airflow, Dagster, or Azure Data Factory)
Experience building on cloud-native data platforms and lake / lakehouse / warehouse architectures,



with layered (medallion-style) modeling
Hands-on experience preparing data for AI/ML or analytical consumers — embeddings / vector pipelines, RAG-/feature-ready datasets, or equivalent — including deduplication, normalization, and validation
Familiarity with vector search and embeddings in production (MongoDB Atlas Vector Search or equivalent)
Demonstrated use of AI-assisted development tools (Claude Code, Copilot, Cursor) for data and pipeline work
Strong grasp of data quality, testing, lineage, and pipeline observability practices
Comfortable working in a complex, specialized domain. MEP / AEC / construction experience is a plus; appetite to learn the domain is required
It's a plus:
Experience with the Azure data ecosystem (Data Factory, Synapse Analytics, Azure Functions, Event Grid)
Lakehouse platforms (Databricks, Snowflake) or open table formats (Iceberg, Delta, Hudi); feature stores (Feast or equivalent)
Streaming / event-driven data processing (Kafka, Event Hubs, Spark Structured Streaming)
CDC and cross-engine sync (MongoDB Change Streams, Debezium, or equivalent)
Experience with geometric / BIM / CAD data or other multi-modal, unstructured source data
Knowledge-graph, ontology, or semantic-layer exposure
Data governance for AI/agent access to production data: query-cost controls, read-path safety, lineage, audit
SOC 2 and data-classification awareness
This call is made within the framework of Law ****** on the Promotion of Employment for Persons with Disabilities, including individuals registered in the National Registry of Persons with Disabilities of the Ministry of Social Development

📌 Senior Data Engineer (Buenos Aires)
🏢 Marvik
📍 Buenos Aires