Senior / Lead Platform Forward Deployed Engineer - Cogentiq I2C

• Own the deployment architecture of client installations from environment readiness through hypercare: bundle strategy, environment topology (Dev, QA, production; per-client catalogs), release sequencing, and sign-off that the installation meets the validation standard.
• Navigate Unity Catalog and platform governance Client and internal IT control catalog and metastore provisioning, storage decisions, and access models. You will assess what their governance model means for our setup scripts, negotiate provisioning paths, and know when a shared-infrastructure convenience creates an SOW, NDA, or data-isolation problem that must be escalated rather than worked around.
• Set and enforce the platform engineering standards: idempotent, rerunnable deployments; fail-fast error handling with no silent except-and-continue paths; environment-variable-driven parameterisation with a single point of change per client value; and configuration conventions that hold across Databricks today and Fabric tomorrow.
• Own the shared utilities architecture: interface-function contracts and factory-pattern abstractions across Databricks, Fabric, PostgreSQL, SQL Server, and storage back ends, packaged as versioned wheel files through Azure Artifactory, so swapping a client's platform is a configuration change, not a rewrite.
• Make the hard ingestion calls: auto-loader versus custom conflict-driven ingestion when the client will not provision Databricks, checkpoint and reprocessing strategy, retry and alerting design per pipeline, and writeback reliability to the application database.
• Run incident command during deployment and hypercare. When a deployment breaks in a client environment, you run the incident: triage across the data estate, the application database, and upstream ERP feeds; communicate honestly with the client; and land the fix and the post-incident correction to the playbook.
• Own the deployment playbook as a product. The quick-start guides, validation suites, and synthetic data configurations you leave behind must make the next deployment cheaper. The bar is that a newcomer deploys the platform from your playbook with one setup script.
• Coach junior Platform FDEs, reviewing their work against the standards you set and growing them toward deployment ownership.

• 6 or more years in data platform or deployment engineering, with at least two years deploying into environments you did not control (client tenancies, regulated environments, or equivalent).
• Deep Azure Databricks: Asset Bundles, workflows, Unity Catalog and metastore governance, cluster and library management, service-principal automation, and the CLI.
• Strong Python and SQL, with the judgement to review pipeline code for idempotency, failure behaviour, and configuration hygiene, not just correctness.
• PostgreSQL in production: schema evolution, write reliability, connection behaviour under load.
• CI/CD for data platforms: Git-based release flows, semantic versioning, artefact repositories, and backward compatibility across component versions.
• Domain literacy in order-to-cash and accounts receivable data: invoices, receipts, remittances, customer masters, and ERP AR structures in Dynamics 365 or SAP. You cannot validate a deployment whose data you do not understand.
• Client-facing composure: you will be the technical face of the deployment to client IT and finance stakeholders.
• Nice to have: Microsoft Fabric, observability design (OpenTelemetry, Application Insights), and experience taking a data product through pilots at multiple clients in parallel.

• Two client deployments taken from environment readiness through hypercare, each validated against the standard suite and signed off on schedule.
• A deployment playbook that demonstrably reduced the cost of the deployment that followed it, measured in elapsed time from access granted to validated installation.
• Platform standards (idempotency, fail-fast, parameterisation) adopted across the data engineering codebase, evidenced in review practice, not just documentation.
• At least one governance or infrastructure negotiation with client IT resolved without escalation to leadership.

If you like wild growth and working with happy, enthusiastic over-achievers, you'll enjoy your career with us!

Not the right fit? Let us know you're interested in a future opportunity by clicking Introduce Yourself in the top-right corner of the page or create an account to set up email alerts as new job postings become available that meet your interest!

It's fun to work in a company where people truly BELIEVE in what they are doing!

We're committed to bringing passion and customer focus to the business.

Cogentiq I2C is Fractal's agentic AI platform for invoice-to-cash operations, spanning Collections, Cash Application, Deductions, Credit Risk, and Invoice Management. The platform runs on a four-tier architecture: a Next.js application layer, FastAPI services, the Cogentiq APA agentic orchestration layer, and a data estate built on Azure Databricks and PostgreSQL. Deployments land in client Azure tenancies against live ERP data (Dynamics 365, SAP), which means deployment engineering is a first-class discipline, not an afterthought to development.

Why this role, now

The platform has crossed from being built to being deployed. Collections is entering pilot with enterprise clients, and the deployment machinery is real: Databricks Asset Bundles with one-click, script-based deployment; auto-loader ingestion running historical and incremental loads into a common data model; versioned bundles and wheel-file packages delivered through Git and Azure Artifactory; and a quick-start standard under which a newcomer must be able to stand up the full data platform by running one setup script.

This role carries that machinery into client environments and makes it hold under client governance, client data, and client timelines. It owns everything between our release artefacts and a running, observable, correctly governed installation; candidates who want to work on agent behaviour and LLM integration should apply to the AI FDE track instead.