Solution design · Jan 2024 – Sep 2024 · Palantir Foundry

CRM Application Development using Palantir Foundry

A comprehensive CRM built on Foundry’s full application stack — Pipeline Builder, Ontology Manager, Workshop, Automate, Vertex, Code Workspaces, and AIP — connecting company and people data to automated client onboarding and real-time client–employee management.

company & people data→cleanse & enrich→Pipeline Builder→ontology objects→Automate assignment→Workshop CRM ↩ AIP Logic / LLM generates synthetic data and specialized table splits for development and demos
Full stack
Foundry suite
AIP + LLM
Synthetic data
Auto-assign
By state & industry
Workshop
Real-time CRM UI
00

Overview

product intent

The CRM consolidates company and people datasets into a governed ontology, automates employee-to-client assignment, and surfaces day-to-day relationship management in Workshop — with metrics, filters, overlays, and custom widgets for real-time insight.

Two design choices define the build. First, data quality is treated as a first-class pipeline stage — analysis and cleansing happen in Jupyter Code Workspaces and SQL Workbooks before Ontology load. Second, onboarding automation lives in TypeScript Automate functions, so assignment policy (state + industry) can evolve without redeploying the Workshop UI.

SOURCES & WORKSPACES Companies People AIP / LLM synthetic Jupyter + SQL cleanse PIPELINE BUILDER — ETL Ingest & transform Specialized tables Quality gates Load to Ontology ONTOLOGY Client object set Employee object set Assignment action types AUTOMATE TS auto-assign state + industry Onboarding flow new client trigger WORKSHOP + VERTEX Filters & widgets Overlays & vars Metrics / KPIs Client–employee UI Vertex graphs
End-to-end CRM lineage from source data through ontology automation to Workshop
01

Pipelines & Data Quality

Pipeline Builder · Python · SQL

ETL backbone

Pipeline Builder

Automated ingestion, transformation, and loading across CRM components

Ingestion
Company and people datasets landed and normalized for downstream CRM tables.
Analysis & cleansing
Python in Jupyter Code Workspaces and SQL in Code Workbooks improve structure, fix quality issues, and prepare Ontology-ready frames.
Specialized tables
Curated splits feed distinct CRM concerns (accounts, contacts, assignments, activity) without overloading a single wide table.
Integration
High data quality gates before Ontology load so Workshop and Automate operate on consistent entities.
02

AIP & Synthetic Data

LLM-assisted enrichment

Custom AIP Logic and LLM integration generate synthetic data where production samples are thin — accelerating development, demos, and edge-case coverage without waiting on upstream systems.

AIP Logic
Custom logic drives synthetic generation and table specialization inside Pipeline Builder.
LLM integration
Language models assist structured synthetic output aligned to CRM schemas.
Downstream benefit
Richer fixtures improve Ontology modelling and Workshop UX validation before full production load.
GuardrailSynthetic paths are isolated from production writeback; Ontology Action Types remain the sole writers of operational CRM state.
03

Ontology Model

objects · actions · links

CRM semantic layer

Ontology Manager
Object Sets
Clients, employees, and supporting CRM entities modelled as first-class Object Types with queryable properties.
Action Types
Governed writes for onboarding, assignment, and relationship updates — auditable and permission-aware.
Link Types
Relationships between companies, people, and assignees enabling Workshop navigation and Automate targeting.
SQL manipulation
SQL used for data manipulation and querying during curation and validation prior to object sync.
04

Automate & TypeScript

onboarding automation

Auto-assignment algorithm

TypeScript Functions

Employees assigned to new clients based on state and industry

Trigger
New client creation enters Automate, invoking TypeScript assignment logic.
Matching policy
State and industry attributes select the best employee assignee — reducing manual routing.
Writeback
Assignment persisted via Ontology Action Types so Workshop immediately reflects ownership.
Extensibility
Additional Automate functions layer follow-on onboarding steps without redesigning the UI.
05

Workshop Experience

client–employee management

Application interface

Workshop · Vertex
Visualizations
Charts and relationship views for portfolio health and coverage.
Filters & variables
Interactive slicing across clients, regions, industries, and assignees.
Custom widgets & overlays
Focused UX for create/edit flows and contextual detail without leaving the CRM canvas.
Functions & metrics
Real-time KPIs and Ontology Functions wired into the UI for live operational insight.
Vertex
Graph-oriented exploration of company–people–employee relationships.
06

Delivery Stack

tools & outcomes
Pipeline Builder
ETL
Ingestion, transforms, specialized tables, quality before Ontology load.
Ontology Manager
Semantics
Object Sets, Action Types, and links for CRM entities.
Automate + TS
Automation
State/industry auto-assignment for new client onboarding.
Workshop + AIP
Experience
Real-time CRM UI with LLM-assisted synthetic data support.

Environment

stack
Platform
Palantir Foundry — Code Workspaces, Ontology Manager, Workshop, Pipeline Builder, Automate, Vertex, AIP
Languages
TypeScript, Python, SQL
Period
January 2024 – September 2024