Field analysis

The consulting half of the job

Short briefs from the APPLY track of my FDE program: reverse-engineering how teams at Palantir, Stripe, Ramp, and others scope and deliver AI systems in the field. Written as I work through each case — analysis in my own words, not summaries of theirs.

FA-001drops in month 1

Palantir · Systems thinking

How Foundry maps enterprise data flows across silos

Analysis of how system dependency graphs let forward deployed teams reason about siloed departments as one system — and why the map itself becomes the client deliverable.

›Dependency graphs surface failure cascades before they happen
›The FDE's first product is often a shared picture, not code

FA-002drops in month 1

Stripe · Problem structuring

MECE decomposition of payment fraud detection

Working through how a monolithic problem — 'detect fraud' — breaks into 12 mutually exclusive sub-problems that can be scoped, prioritized, and shipped independently.

›Decomposition quality determines everything downstream
›Sub-problems map one-to-one to shippable increments

FA-003drops in month 1

Ramp · FDE delivery

Scoping an expense anomaly detector: client call to production

Reverse-engineering how Ramp's forward deployed engineers turned an initial client call into a scoped, deployed anomaly detection system.

›Discovery questions decide the architecture more than technology choices do
›PoV → PoC → MVP stage gates keep scope honest

FA-004drops in month 2

Salesforce · Data foundations

Unifying CRM data from 7 sources for AI enrichment

How disparate CRM sources become a single customer view fit for AI — and what 'fit for AI' actually demands of data quality.

›Entity resolution is the hidden 80% of the work
›Data contracts beat data heroics