Final SOW (Artificial Intelligence)-Redacted.pdf

STATEMENT OF WORK

FOR

Artificial Intelligence Course

Prepared by Studies and Analysis Squadron (SAS)

JBSA-Randolph AFB, Texas 78150

Prepared for

DEPARTMENT OF THE AIR FORCE

1. GENERAL INFORMATION.

1.1. REQUIREMENTS FROM CONTRACTOR

The Contractor shall provide all information, lectures, and coding packages necessary to conduct training in artificial intelligence and its relationship to operations research (data pre-processing to analysis and reporting), incorporating all aspects of data science process to operations research problem unique to the mission of HQ AETC SAS at JBSA-Randolph AFB, Texas 78150.

Instructors shall provide code files for lessons that ensure that students are able to quickly use tools discussed in lesson for activities. The instructors shall have a PhD in Data Science or Artificial Intelligence (AI) or have experience teaching classes in Data Science or AI.

1.2. AUDIENCE BACKGROUND

The audience is composed of analysts that have graduate degrees and bachelor’s degrees in fields such as Operations Research, Applied Math, and Computer Science. The audience will have experience with coding and will be familiar with artificial intelligence concepts. All participants in the course will have at least a working experience with Python.

1.3. REQUIREMENTS FROM HQ AETC SAS

The instruction will be in-person at Randolph AFB at the facility provided by HQ AETC SAS.

HQ AETC SAS will provide the classroom, a large screen for presentations, student laptops, and internet for the purpose of conducting this course. HQ AETC SAS will provide a survey of student skills for the contractor to tailor lessons to fit student needs and desires in learning objectives.

2. DESCRIPTION OF SERVICES.

All activities provided shall be coded in Python. For activities, the contractor shall provide code files for students to work on that prepare for immediate hands-on use of topic of lesson. For example, in a lesson on Limited Language Models (LLMs), the instructors shall distribute code files after the lesson so students can run a LLM on their laptop immediately after.

The planned time allocation ratio between lessons and activities will be 1:1, with one hour of lessons followed by one hour of hands-on learning. Short breaks will be administered between lessons. Prior to rendering services, the contractor shall tailor specific course content to current needs of students. HQ AETC SAS may request the contractor to emphasize certain lesson topics based on needs of students.

The contractor shall perform:

2.1.1. Overview and Working Environments set up

2.1.2. Lecture and hands-on coding training on Probability Theory, Probability

Distributions and First Order Statistics

2.1.2.1. Lesson includes but not limited to: most common distributions and assumptions, statistical tests for distributions, and characteristics of datasets.

2.1.2.2. Hands-on training includes but not limited to: Identifying distributions, identifying assumptions about distributions and what types of tests can be used.

2.1.3. Lecture and hands-on coding training of Probability in Data Science

2.1.3.1. Lesson includes but not limited to: concepts such as Bayes’ Theorem, conditional probability, as well as examples for each concept taught.

2.1.3.2. Hands-on training includes but not limited to: use of Numpy and pandas for exercises with an example dataset.

2.1.4. Lecture and hands-on coding training of Data Types, Collection, and Cleaning

2.1.4.1. Lesson includes but not limited to: Teach on all different data types, best use cases, examples for use.

2.1.4.2. Hands-on training includes but not limited to:

2.1.5. Lecture and hands-on coding training of Data Transformations and Feature

Engineering

2.1.5.1. Lesson includes but not limited to: teaching on relevant features to train a model, how to identify from example problems, the concept of noise in a model, and different techniques that can be used in various problems.

2.1.5.2. Hands-on training includes but not limited to: Developing a model and identifying relevant factors and opportunities to practice backwards selection

2.1.6. Lecture and hands-on coding training of Optimization and Markov Decision Processing

2.1.6.1. Lesson includes but not limited to: Mathematical concepts and theory of MDP, real world examples, and relevant to US Air Force. Optimization models and examples/use cases.

2.1.6.2. Hands-on training includes but not limited to: how to develop a MDP model and an example of achieving optimization.

2.1.7. Lecture and hands-on coding training of Machine Learning Algorithms and Applications

2.1.7.1. Lesson includes but not limited to: supervised and unsupervised learning, image recognition, structured and unstructured data.

2.1.7.2. Hands-on training includes but not limited to:

2.1.8. Lecture and hands-on coding training of Classification and Regression

2.1.8.1. Lesson includes but not limited to: logistic and linear regression, K-nearest neighbors, decision trees. Assumptions, limits, and use cases for each.

2.1.8.2. Hands-on training includes but not limited to: running regression and interpreting results.

2.1.9. Lecture and hands-on coding training of Support Vector Machines and Modeling techniques

2.1.9.1.1. Lesson includes but not limited to: advantages, disadvantages, use cases and core concepts of technique

2.1.9.1.2. Hands-on training includes but not limited to: running and interpreting results of technique

2.1.10. Lecture and hands-on coding training of Decision Trees, Neural Networks, Deep

Learning and Transformers for LLMs

2.1.10.1. Lesson includes but not limited to: core concepts, mathematical foundations, and use cases

2.1.11. Lecture and hands-on coding training of Reinforcement Learning

2.1.11.1. Lesson includes but not limited to: core concepts, mathematical foundations, and use cases

2.1.11.2. Hands-on training includes but not limited to: use of reinforcement learning

2.1.12. Lecture and hands-on coding training of Explainable AI and XAI Models (LIME, SHAP)

2.1.13. Perform Practical Exercise incorporating all aspects of the Data Science Process to an Operations Research problem, maybe one of their own if coordinated.

2.1.13.1. Hands-on training includes but not limited to: one final project that will take at least the last day for development and presentation, incorporating all of lessons into solving one large problem.

Contractor shall combine lessons as see fit with permission of AETC HQ SAS

3. GENERAL INFORMATION.

3.1. OBJECTIVE. The objective of this training is for attending analysts to learn about AI tools and most up-to-date best practices in the disciplines of Data Science and Artificial Intelligence, as well as develop a working knowledge of the tools taught by the contactor.

Analysts will be able to use all of the lessons in a large case study where all/most of the discussed concepts are brought together. The contractor shall provide students with a certificate acknowledging participation in the course and completion of final project.

3.2. SALIENT CHARACTERISTICS:

Two or more instructors shall have a PhD in Data Science or Artificial Intelligence.

Instructors shall have teaching experience at an accredited university in a subject matter of this course and publications associated with AI and data science.

Instructors shall have experience in teaching as it refers to military studies.

Instructors shall provide code files for students to work in that prepare for immediate exploration of topic of lesson.

The planned time allocation ratio between lessons and hand-on coding will be 1:1.

Based on conversations with HQ AETC SAS and the contractor, the time spent on the lesson objectives shall emphasize some and minimize others.

3.3. QUALITY CONTROL. The contractor shall be responsible for quality control for all work accomplished during the performance of the task. The contractor shall provide and maintain a Quality Control Plan (QCP) which ensures the requirements in this PWS are met.

The QCP will be considered acceptable and approved unless the contractor is notified by the Contracting Officer (CO) prior to award. The QCP shall be included in the technical proposal for evaluation and acceptance. As a minimum, the contractor shall develop quality control procedures that address the areas identified in section 2, Service Summary.

3.4. QUALITY ASSURANCE. The government will periodically evaluate the contractor’s performance by appointing a Contracting Officer Representative (COR)(s) to monitor performance to ensure services are received. The COR will evaluate the contractor’s performance in accordance with the Quality Assurance Surveillance Plan (QASP) for this requirement. The contractor shall provide lesson slides 5 business day prior to arrival to Randolph AFB provided facility.