DataRobot_JA_(signed)_Redacted.pdf

UNITED STATES MARINE CORPS

MARINE CORPS INSTALLATIONS COMMAND

3000 MARINE CORPS PENTAGON

WASHINGTON, DC 20350-3000

J&A No. 19-009

V1.0, 13 June 2019

JUSTIFICATION AND APPROVAL (J&A)

FOR USE OF BRAND NAME OR EQUAL DESCRIPTIONS, OR PROPRIETARY SPECIFICATIONS AND

STANDARDS, IN SOLICITATIONS

1. Contracting Activity

The Contracting Activity is the Marine Corps Installation Command (MCICOM) Headquarters (HQ) Contracting located in Arlington, VA.

2. Description of the Action Being Approved

This action being approved is justification for other than full and open competition of a Brand Name requirement for a

DataRobot software license in support of the Marine Corps Force Preservation Directorate (FPD). This requirement will be awarded as a Firm-Fixed Price contract to an authorized reseller of DataRobot software, the Original Equipment

Manufacturer.

3. Description of the Supplies/Services

This requirement provides for license access to the DataRobot Automated Machine Learning (AutoML) software in support of FPD. FPD requires AutoML support to enable senior leaders to focus on better outcomes by revolutionizing how enterprise analytics is integrated with cross functional research. This requires a multi-disciplinary approach in bringing together expertise from both social science and applied mathematics in order to: define research questions from motivating problems; gather the appropriate enterprise data; explore the data; and build and refine analytic models before providing input for resourcing decisions. The Period of Performance (PoP) is as follows:

Period of Performance Dates

Base 30 September 2019 - 29 September 2020

Option 1 30 September 2020 - 29 September 2021

Option 2 30 September 2021 - 29 September 2022

In order to increase the capabilities and analytic capacity, FPD seeks to leverage cutting-edge techniques of applied mathematics through the use of AutoML tools, which streamline several facets of the machine learning process that ordinarily require highly trained data scientist. The particular phases of the machine learning and data science application which must be streamlined include: feature selection; feature preprocessing; feature engineering; model training and selection; parameter optimization; and accompanying technical support. The table below lists the materials required for this effort:

Description Qty Part Numbers

Modeling Workers, Reserved Prediction Cores; AutoML (Base Edition) 1 DR-BSM-ML-12-8-11

Enabled Success Package 1 ES-PKG

The acquisition will be funded through Fiscal Year (FY) 2019 Operation & Maintenance Marine Corps (OMMC) funds.

4. Statutory Authority Permitting Other Than Full and Open Competition

This document meets the regulatory requirement at DFARS 206.302-1(c)(2) to execute the justification and approval addressed at FAR 6.303 in order to use a DataRobot AutoML license. FAR Subpart 13.5 procedures are being utilized for this acquisition in accordance with 41 U.S.C. 1901 or the authority of 41 U.S.C. 1903.

3000 MARINE CORPS PENTAGON

WASHINGTON, DC 20350-3000

J&A No. 19-009

V1.0, 13 June 2019

5. Rationale Justifying Use of Cited Statutory Authority

FPD employs a cross functional team of analysts to examine complex data sets and provide senior leaders with recommendations to resolve complex problems. Currently, the FPD team uses software tools coded in languages such as

R and Python as well as proprietary software tools such as Tableau and Linux to gather, analyze and refine data. These tools require a data analyst to build models and to manually increase capabilities. Unfortunately, increased demand for data analysts both in the private sector and DoD have resulted in a critical shortage of trained and skilled data scientists.

The keen competition has resulted in high turnover and inability to quickly fill up positions. In an effort to leverage emerging technology, FPD sees AutoML solutions as a way to both enhance the current analytic team and mitigate the impacts of turnover.

AutoML is the process of automating end-to-end the process of applying machine learning to real-world problems. In a typical machine learning application, practitioners must apply appropriate data pre-processing, feature engineering, feature extraction, and selection methods to make data for software tools amenable for machine learning. AutoML is an artificial intelligence based solution that automates the Machine Learning process by producing simpler solutions and models that are usually designed by hand. There are three types of AutoML capabilities: open source, closed source, and standard. Only the closed source type works in FPD’s network. Although there are multiple closed source AutoML solutions in the market, DataRobot’s AutoML solution is the only one compatible with FPD’s closed system architecture, which is a closed network that does not work with the Cloud. FPD currently lacks an approved DoD cloud computing environment capable of holding Personally Identifiable Information (PII) with the necessary bandwidth to conduct analytic computing.

In addition, DataRobot AutoML is the only solution that provides a suite of tools that fully maximizes FPD’s team performance. In depth assessment of DataRobot AutoML solution demonstrated an increase in capability via time savings. The time saved by the team in model building was valued at more than $63k, but the increased capacity measured in Full Time Equivalents (FTEs) is estimated at more than $4.4 million annually. Meaning G-10 would have to hire $4.4 million of additional contract FTEs to complete the same amount of work in the same time span. In comparison to other AutoML solutions, the DataRobot AutoML efficiency results from the advance feature engineering capability that requires minimal data analyst input.

6. Description of Efforts Made to Solicit Offers from as Many Offerors as Practicable

Market research was conducted in accordance with FAR Part 10, and revealed that there are multiple authorized resellers of DataRobot AutoML solutions in the commercial market. A Request for Information was posted on FedBizOpps on 05

August 2019 with a closing date of 09 August 2019.

7. Determination of Fair and Reasonable Cost

The Contracting Officer has determined that the anticipated cost to the Government of the supplies covered by this J&A will be fair and reasonable based on competition.

8. Actions to Remove Barriers to Future Competition

If FPD transitions over to an approved DoD cloud computing environment, a brand name requirement might not be necessary in the future as there are multiple cloud AutoML solutions in the commercial market. The contracting officer will reassess this requirement during follow-on acquisition planning.

3000 MARINE CORPS PENTAGON

WASHINGTON, DC 20350-3000

J&A No. 19-009

V1.0, 13 June 2019

CERTIFICATIONS AND APPROVAL

TECHNICAL/REQUIREMENTS CERTIFICATION

I certify that the facts and representations under my cognizance which are included in this Justification and its supporting acquisition planning documents, except as noted herein are complete and accurate to the best of my knowledge and belief.

Technical Cognizance:

_____________________ Major Gregory T. Zerr __703-604-6343______ _05 August 2019____

Name (Printed) Phone No. Date

LEGAL SUFFICIENCY REVIEW

I have determined that this Justification is legally sufficient.

Signature Name (Printed) Phone No. Date

CONTRACTING OFFICER CERTIFICATION

I certify that this Justification is accurate and complete to the best of my knowledge and belief.

Signature Name (Printed) Phone No. Date

Ellen Brandau Clark 703-614-0070 August 16, 2019

F. PAUL LIRIO 703-604-3236 21 AUG 2019