bomonike

palantir.png How to learn Palantir system for enterprise use.

Overview

Why?

Many are taking time to learn Palantir because Palantir provides free tutorials and cloud time.

A “Palantir” is the mythical artifact from the “Lord of the Rings” which gives one the power of seeing and understanding the world.

Technically, at its core, Palantir provides an object-based database they call an Ontology. It’s unique in that branches of data in it can be versioned (forked) like GitHub source code. That enables “what-if” changes to be analyzed, which inform decisions. When a satisfactory outcome is achieved, changes can be “materialized” into production data. In addition to master reference data and transactional data, a Palantir Ontology can store decisions the organization makes over time.

Palantir provides tools to create:

REMEMBER: The Ontology has a different vocabulary:

QUESTION: How does Palantir identify and issue alerts based on events and metrics triggers?

Palantir’s tools were not hobled together through acquisitions over time. Palantir being built initially for military use required its data governance features to be built-in (to include object lineage time series, auditability, and traceability). So Palantirians have a better chance of getting apps working together in production-scale – in SaaS and on-premise.

palantir-processes.png

PROTIP: What’s not mentioned in Palantir’s current docs are “AI agents” that have the “agency” to take autonomous action on their own, without human intervention, based on alerts triggered by events.

Applied Notational Project

  1. Play step-by-step video by Gena Coblentz at Ontologize

    1. Begin with raw data sets
    2. Prep data sets for ontology
    3. Combine into workshop app with write back

  2. End-to-end speedrun https://learn.palantir.com/speedrun-your-first-e2e-workflow

    https://learn.palantir.com/deep-dive-building-your-first-pipeline Deep Dive: Building your First Pipeline

    https://learn.palantir.com/deep-dive-creating-your-first-ontology Deep Dive: Creating your First Ontology

    https://learn.palantir.com/deep-dive-building-your-first-application Deep Dive: Building your First Application

  3. Sign in to the lesson text at:

    https://learn.palantir.com/applied-notional-project-automobile-inbox

    palantir-notational-proj-2862x1746.png

    Expected Completion Time: This course will take a Foundry novice (1 week of experience) around 1 full business day to complete. If you are someone who has only completed the End to End Speedrun, it may take longer, but that’s okay! This project is designed to challenge you so that you come out with all the skills needed to develop on your own in the real world!

    Overview: In this course, you will take on the role of a Foundry engineer tasked with constructing a dashboard for a used car leasing company. Designed to mimic real-world scenarios, the course provides raw datasets and application requirements, but not step-by-step instructions, challenging you to develop your problem-solving and development skills.

    The application requirements are framed from the viewpoint of a car leasing manager, who needs assistance in monitoring his fleet and addressing related issues. By the end of the course, you will have created a fully functional application that enables a company to oversee their automobile fleet, examine their cars, and rectify issues.

    This course is an excellent opportunity to learn Foundry development without rails, allowing you to gain practical skills in decomposing and solving problems. You will learn to critically analyze application requirements and transform them into development tasks. This hands-on experience will equip you with the ability to perceive Foundry not as a single-purpose application, but as a suite of tools designed to work together, ultimately helping you become an independent, self-sufficient developer.

    [1:21] Prerequisites Permissions & Help Project Objectives and Requirements Project Summary

  4. From within the course:

  5. Permissions to capabilities:

    • Project Creation [Optional, can use an existing project instead]
    • Raw Data Upload
    • Object Type Creation in the Ontology
    • Link Type Creation in the Ontology
    • Action Type Creation in the Ontology
    • Workshop Application Creation

    In almost all cases, your necessary permissions can be boiled down to having edit access in the project you are working in, as well as the ability to create and edit ontology object types.

  6. [1:29] Getting Help: https://www.palantir.com/docs/foundry

    AIP prompts

  7. Project Summary:

    I own a car leasing company and I am looking to make an application to support my business. Recently, my customers have been reporting lots of issues with their cars. I need a system to help me keep track of this! I would like you to create an application in Foundry for me and my employees to use.

    In the first tab, I would like a high level dashboard of my cars. I have provided you with raw data that should have everything you need to build this application.

    In the second tab, I would like a list of outstanding issues on my cars. In addition to viewing details about existing issues, I want my employees to be able to update the priority of an issue, and mark an issue as closed.

    Keep in mind, the raw data is a bit messy, so you might have to clean it up before it can be application ready!

  8. [1:34] Tips Before Getting Started

    The only Foundry applications that you need to use in order to complete this project are Pipeline Builder, Ontology Manger Application (OMA), and Workshop.

  9. Raw Data Downloads:

    Car info raw.csv Model information raw.csv Issues raw.csv Historical maintenance log raw.csv

    1. cars dataset called Car info raw. This dataset contains basic information about each of the cars.
    2. maintenance history dataset called Historical maintenance log raw. This dataset contains historical information on maintenance performed on the cars. I don’t need row level information from this dataset, I instead want aggregated information from this dataset, such as number of number of maintenance events per car.
    3. Model information raw. This dataset provides additional information on a car based on the manufacturer and model of that car.
    4. Issues raw. This dataset is a running list of active issues on the cars.

    Objective 1 - Data Cleaning Objective 2 - Cars Homepage Objective 3 - Issues Inbox Hints Hints Overview & General FAQ Architecture Raw Datasets Pipeline Builder Ontology Manager Application Workshop

The Ontology

https://www.palantir.com/docs/foundry/ontology/overview/

Palantir’s uses the word “Ontology” instead of “database” to emphasize that, in addition to holding data like SQL databases, it maintains what’s necessary to make a digital twin modeling the dynamic interactions within organizations.

Built for each project are custom and pre-built
“semantic” elements:

palantir-objects-702x81.png

“kinetic” elements:

For example, the “Assign Employee” Action type would define how users can change the role property value for a given Employee object. This Action type could require a parameter definition to ensure input of the new role in a standardized form and can include rules for how to automatically create a link between the Employee object and that of a new Manager. Also, include a notification side effect that will notify the old and new manager of the change. Plus, Validate that authorized employees such as those working in human resources can perform the Action.

An object is an instance of an object type.

types of data and links:

Interfaces describes the shape of an object type and its capabilities. Interfaces provide object type polymorphism, allowing for consistent modeling of and interaction with object types that share a common shape.

The Ontology regarding decisions:

palantir-role-actions.png

That enables Palantir to conduct predicate-based data mining.

treats data as code: Unlike SQL databases, experiments to data are a built-in feature. Copies of Palantir’s Ontology

VIDEO Fondary is fast!

Integration also applies to the granular access allowed human users and automated actions depending on specific role and rights based on object and links within the Ontology.

Palantir provides a full suite (stack) of tools instead of others (but also integrates with them). VIDEO “Workshop App: Supply Chain Control Tower” provides a quick tour:

For “full spectrum” enterprise AI:

Operational App Building:

https://www.youtube.com/shorts/Qvh2DovBi64 Parady: “We’re making more money”

https://www.youtube.com/watch?v=A0yim6QYiUo&pp=0gcJCX4JAYcqIYzv “What does Palantir Actually do?” by CTO Shyam Sankar on Shawn Ryan

Dive right in

Get hands-on.

Out-of-the-box workflow templates or reusable workflow components to create workflow applications quickly, often within a matter of days. This accelerated development process ensures that organizations can realize the value of their data and insights swiftly, enhancing efficiency and responsiveness.

  1. At https://build.palantir.com click “Sign Up” to get an environment using your company email.
  2. At https://learn.palantir.com click “Start Building Now” to get an dev account using your email, phone, credit card, etc.

  3. https://accounts.skilljar.com/accounts/signup/ with a company email.

    WARNING: I can’t find what Palantir charges for dev & prod use.

  4. https://learn.palantir.com/speedrun-your-first-e2e-workflow at https://github.com/palantir/first-e2e-workflow

  5. https://learn.palantir.com/foundry-aip-builder-foundations-quiz 10 questions

  6. Deep Dive: Data Protection Tools in Foundry

  7. https://learn.palantir.com/applied-notional-project-automobile-inbo
  8. Applied Notional Project VIDEO to build a dashboard for a used car leasing company with minimal step-by-step instructions based on requirements taken from a notional statement of work. [1-2 days] an operational workflow

Socials

  1. On your calendar, set a recurring WEEKLY appointment to to read Announcements and Release Notes from Palantir about specifics about changes that impact how you build software.

  2. Accept CodeStrap.me Discord invite

As of this writing, the AIP con occured on June 6, 2024. VIDEO “Our Ontology is “the only thing that creates quantifiable transformational value in Enterprise.” Deep Dive: Creating Your First Ontology

The company

Palantir was founded May 3, 2003 after 9/11 to serve CIA and other law enforcement companies.

In early 2025 Palantir (stock PLTR) was among the fastest-growing among all stocks, peaking at $125 before the Trump tariff meltdown brougt it back to $80/share at P/E of 1457.

Co-founders are billionaire Peter Thiel, Joe Lonsdale, Nathan Gettings, Stephen Cohen, and Alex Karp.

They came from PayPal which had to develop complex fraud detection software to stem large losses with “Pattern of Life Analysis” that later was used to hunt terrorists: VIDEO.

CEO Alex Karp wrote NYTimes Bestseller “The Technological Republic” “Hard Power, Soft Belief, and the Future of the West”. VIDEO. He was named by The Economist’s “best CEO of 2024”. But he has a (middling) 69% approval rating on Glassdoor. Karp describes himself as “anti-woke, pro-Israel, pro-Ukraine”. VIDEOS:

VIDEO: Fund manager Kathy Wood said “Alex Karp believes that Plantir will be the largest AI company in the world. .. I think that might be right. … It could be a big disruptor to Microsoft.” Others:

Glassdoor’s employee rating of the company is 3.8/5.0 (on the lower end of scale, not among Best Places to Work). One employee commented:

LinkedIn reports 2.7 year average tenure (short vs. other employers)

Palantir employees are called Palantirians.

The company has “we don’t have salespeople” philosophy. “each contract is tailored to the customer’s needs.” One employee commented on Indeed: “They will do whatever they possibly can to avoid paying out owed commissions. Every salesperson I have worked with at Palantir have been owed hundreds of thousands in commissions.”

Corporate offices moved from the Silicon Valley to Denver, CO.

Jobs listed in Palantir careers and LinkedIn show “Forward Deployed Software Engineers” (Deltas), what others call Solution Engineers) around the world and within engineering offices in Seattle, DC, NYC:

Tech Stack

Palantir sells its data operations platform as “Gotham” to governments and “Foundry” to commercial clients.

  1. Click “INSTALL” at “Ontology SDK (OSDK) With AIP Logic - Build a To Do application Powered by AIP Logic

  2. Enter your enrollment URL such as “almond-latte.palantirfoundry.com”.

    QUESTION: Where do we get the enrollment URL???

  3. Configure application access

The Developer Console is the platform used to create both Ontology SDK applications as well as OAuth clients (formerly known as third-party applications in Control Panel). To access the Developer Console, open the application portal and search for developer console.

Object Views are a central hub for all information and workflows related to a particular object. This includes key “biographical data” about an object, any linked objects, key related metrics, and links to (or embedding of) key analyses, dashboards, and applications related to the object.

Use Jupyter to analyze data from the Ontology or trigger AIP Logic functions

Create a Dataset of tasks within projects within the Ontology

AI/ML

To operationalize AI/ML in Foundry for live inference with the Ontology:

  1. Create a LLM model in Foundry.
  2. Configure a direct model deployment.
  3. Publish a simple wrapper function for your model and optionally call it from another function to orchestrate complex logic around your model.
  4. Use that function for live inference in Workshop, Vertex and other end-user facing applications.

https://www.palantir.com/docs/foundry/model-integration/tutorial-train-code-repositories/#2b4-how-to-test-inference-logic-in-code-repositories Ontology Objects can also be backed with datasets that leverage a model for batch inference -

Functions

A Function is a piece of code-based logic that takes in input parameters and returns an output. Functions are natively integrated with the Ontology: they can take objects and object sets as input, read property values of objects, and be used across action types and applications that build on the Ontology.

AIP Logic can be used from anywhere.

AIP Logic functions from front-end React or back-end Python applications.

Call Query functions which uses GPT-4o from a TypeScript Query function.

from which Jupyter notebooks run to analyze data and trigger AIP Logic functions from your OSDK application.

Ontology Software Development Kit (OSDK)

The Ontology SDK is multi-lingual. It supports NPM (Node Package Manager) package for TypeScript, PIP, or Conda for Python, Maven for Java, and OpenAPI spec for any other language. It generate:

https://www.palantir.com/docs/foundry/ontology-sdk/overview Functions are executed using the Ontology SDK to create a To Do webapp.

QUESTION: Application RID???

The OSDK uses a token that is scoped only to the ontological entities required for a single application to access, in addition to the user’s own permissions to the data.

Palantir platform tooling:

Palantir has three tools for building Apps & Dashboards:

What is

What is Palantir? Eric Timberling’s description

https://build.palantir.com

VIDEO: predicate-based Data mining named for the device in “Lord of the Rings” for seeing vast distances. Rather than using a wide net.

Only AWS cloud - US East or UK.

“AIP” is their product.

Ontology for AI VIDEO

The Ontology serves as the semantic layer that integrates data, models, and business logic, The Ontology is a semantic layer that models a company’s data to ensure it is usable and understandable by all business users. Objects within the Ontology can be linked together to reflect business activities. For example, a signed contract (one object) would be linked to the client it is signed with (another object), which is then associated with cashflows (yet another object). These links create a comprehensive and interconnected view of the business.

The Ontology is composed of objects, which represent unique entities such as banks, cashflows, people, clients, contracts, or policies. These objects serve as the building blocks of the Ontology.

Objects and Links within the Ontology can be configured to granularly provide access rights to different users, depending on their role and rights with regards to data access and modification. It bridges the gap between complex data from various source systems and the business logic, translating it into familiar business terms. This makes the data accessible and actionable for everyone in the organization, providing a single source of truth for data-driven operations and decision-making. https://www.youtube.com/watch?v=Qa5KlRdk5UE

Digital Twin

The Ontology is designed to represent the decisions in an enterprise, not simply the data.

Branch: A branch on the Ontology is a separate version of that Ontology, derived from the main version, designed to enable experimentation and changes without impacting the main branch. This allows users to test and refine adjustments to the Ontology in an isolated environment before merging them back into the main branch.

Proposal: A proposal is analogous to a Pull Request in a version control system, specifically tailored for Ontology branches. A proposal is automatically created alongside a branch and contains metadata such as reviews, name, and descriptions of the changes being merged into the main branch. Proposals serve as a mechanism for reviewing and approving changes made in a separate branch before they are integrated into the main Ontology.

RFC (Request for Comments): A RFC is a document that outlines a new feature or change to the Ontology. It is submitted for review and discussion by the community, and after approval, it is implemented in the Ontology.

Workflows

Foundry workflow DevOps operating system How Palantir Foundry Fuels Your Data Platform

Gotham is for goernments to do decision analysis with AI.

version of Foundary is called “Foundry Gov”

Vertically integrated https://www.youtube.com/watch?v=m2u6KL2Hz2I Workflow Builder

https://github.com/palantir/aip-community-registry

https://www.youtube.com/watch?v=k88WbxMEvPY Palantir Architecture Speedrun | From Integration to Application Palantir

Analytics

With traditional linear, there is often a disconnect between the front-line personnel and the back-office data analytics team. This lack of a feedback loop can hinder continuous improvement and innovation.

Analytics in Foundry go beyond conventional “read-only” paradigms. It can write data back into the Ontology, producing valuable new insights within unified security, lineage, and governance models.

Foundry’s core Analytics applications include:

Certifications

palantir-certs-2048x791.jpeg

  1. At https://learn.palantir.com
  2. Sign up with Palantir Foundry Platform account.
  3. Sign-in

Palantir’s Certification Program was introduced Jan 2022 as part of Palantir’s Amplify service offerings.

https://accounts.skilljar.com/accounts/login/?t=38ttrnhi4ve49&d=1n4nregc1qhy8&next=%2Fauth%2Fendpoint%2Flogin%2Fresult%3Fnext%3D%252F%26d%3D1n4nregc1qhy8 Four cert tracks, $260 for 70% of 60 multiple choice questions in 120 minutes:

Exam guides, including an outline of topics, relevant documentation, and other study materials, will be provided to all candidates.

Code Repositories

VS Code workspaces

Palantir extension for Visual Studio Code

Jupyterlab®

GitHub

Open source from https://github.com/palantir

Certification

All resources and files in the Foundry platform must live in a __.

Dataset x Project - A Project in Foundry is a collaborative space that organizes people, resources, and folders for a particular purpose. All resources (Apps, files, media sets) are stored in Projects while Objects are stored in the Ontology. Ontology Data Lineage

What actions can a user take to load data into Foundry?

Create a data connection from their source system to Foundry and use one of the 300+ connectors available Manually upload files to datasets Manually type their data in Fusion Sheets x All of the above Upload files manually is particularly useful for low volume or for prototyping. Establishing data connection is advised for production workflows. Upload files directly in folders as raw files is not advised, as you won’t be able to consume those files in your pipelines: files need to be uploaded to datasets to be easily consumable.

True or False: Pipeline Builder is the only tool users can use to build production-grade pipelines. True x False Upload files manually is particularly useful for low volume or for prototyping. Establishing data connection is advised for production workflows. Upload files directly in folders as raw files is not advised, as you won’t be able to consume those files in your pipelines: files need to be uploaded to datasets to be easily consumable.

Question 3: Correct answer True or False: Pipeline Builder is the only tool users can use to build production-grade pipelines.

True False Feedback: Other applications are available to create production grade pipelines in Foundry, like Code Repositories; however, Pipeline Builder should always be your go-to starting point.

Which of the following is not needed to create an Object type?

Primary Key Permissions Defined Object properties x Data transformations In Foundry, data transformations are not strictly necessary to create an object type because the ontology system can directly ingest raw data sources.

What can Actions be performed on?

x One or many Objects A pie chart A Workshop application A pipeline Actions are events and edits that are performed on an Object or set of Objects in the Ontology.

True or False: there is no way to restrict who can use an Action.

True x False You can restrict the Action so that only a specific person or group can use it.

True or False: Workshop is the only tool in Foundry to build user-facing applications.

True x False Other applications are available to create user facing applications in Foundry, like Slate or Quiver.

Which of the following is not available in Object Tables?

Properties from an Object Set Function-backed columns Conditional formatting x Action buttons Object Tables can display current object properties, linked object properties, aggregations, object titles, URLs, and Function-backed properties.

Use your final Workshop Application from the Speedrun: Your First End-to-End Workflow course to answer the following question.

How many open orders does ACME Australia have? 0 37 x 5 10

For A4 Paper orders that are open and due in 30 days or less, what is the highest quantity of product included in a single order?

22 x 92 100 6 Utilize the filters, charts, and Object Table in your application to explore your data. (Hint: ensure you use the sorting feature of the Object Table to sort by highest quantity).

v010 + imgs :palantir.md