AI Product Leader (Nanodegree) Certificate (AIPLC)^TM 

Module 0

Python for Product Managers (PMs)

Python is essential for machine learning due to its simplicity and readability, making it accessible for beginners. It supports numerous ML libraries and frameworks like TensorFlow and Scikit-learn, which streamline the development process. Python's strong community and extensive resources facilitate problem-solving and innovation in ML, making it a preferred choice for professionals and researchers in the field.

Outcome:

You will be able to code control flow, lists, loops, functions, strings, dictionaries, classes, and simple algorithms, as well as work with modules and files.  To get there, you will work alongside our instructor through the Python 3 Code Academy course, plus several labs of ours, where you will learn how to troubleshoot some common errors.

Key points:

• 9 hours a week for 3 weeks. However, if students already have some working knowledge of Python we will work with them on more challenging exercises.
• Students may skip this module and begin at Module 1, but you must pass the Python knowledge test.
• Evaluation criteria: Class participation, lab, and written exam.
• Next start date: Thursday, October 24, 2024

Module 1

Statistics for PMs

Statistics underpins machine learning by providing methods to understand and interpret data, which is crucial for designing and training algorithms. It aids in evaluating model performance and understanding results' reliability through techniques like hypothesis testing and A/B testing. By incorporating statistical methods, machine learning practitioners can conduct A/B testing to compare models and make data-driven decisions, ensuring the development of effective and accurate predictive models.

Outcome:

You will be able to create A/B tests and analyze the results to determine if the delta is statistically significant or not. You can articulate the definitions and applications for sample size, power, alpha, 1-sided versus 2-sided tests, and P-values.  You will also be able to implement one-arm bandit and multi-arm bandit tests and be able to calculate NPV, MIRR, and IRR quickly. This module is where we consider the starting point for the AIPLC^TM certificate preparation modules.

Key points:

• 9 hours a week for 2 weeks, including 5 labs and a communications assignment
• Students can opt to challenge this and skip to the next module depending on the strength of their math skills
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Thursday, November 14, 2024

Module 2

Fundamentals of AI/ML for PMs

Outcome:

You understand, at a high-level, what is easy and what is difficult in AI and how to integrate it into an existing product. (Future modules will go orders of magnitude deeper into what is easy versus difficult.)

You will be able to articulate the differences between supervised, unsupervised, and reinforcement learning and select suitable evaluation metrics for all of these. You can also run various algorithms in a Google Colab notebook and a Jupyter notebook hosted by a cloud provider. You will also know why some similar Kaggle competitions selected one evaluation metric over another to lock in this foundation. Additionally, you understand use cases for the top libraries for Python, including Scikit-learn, TensorFlow, Keras, Pandas, NumPy, and up-and-coming libraries such as Fastai. 

Key points:

• 6 hours a week for 1 week, including 5 labs and a communications assignment
• Students can opt to challenge this and skip to the next module depending on the strength of their AI/ML understanding
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Thursday, November 28, 2024

Module 3

Data Ingestion (ETL/ELT), Cleaning & Visualizing for PMs

In machine learning, data Ingestion (ETL/ELT), cleaning, and visualization are crucial for ensuring data quality and usability. ETL/ELT processes efficiently gather and prepare data from diverse sources. Data cleaning removes inaccuracies and inconsistencies, enhancing model accuracy. Visualization aids in understanding data patterns and anomalies, which is essential for insightful feature engineering and algorithm selection. These steps are fundamental for building robust models and achieving reliable, interpretable machine learning outcomes.

Outcome:

 You can create an end-to-end data pipeline in the cloud and can create sophisticated data visualizations with the help of prompt engineering. You are efficient at taking a data set and evaluating it, and cleaning it so it is ready to be used for training. 

Key points:

• 6 hours a week for 3 weeks, including 10 labs where we use plenty of GenAI for code, and a communications assignment
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Thursday, December 5, 2024

Module 4

Machine Learning for PMs

Machine learning is important as it enables systems to learn and improve from experience without being explicitly programmed. It's crucial for analyzing large datasets, identifying patterns, and making decisions with minimal or no human intervention. This technology is applied across various fields, such as finance, healthcare, and marketing, improving efficiency, personalization, and predictive capabilities. By automating complex processes, machine learning drives innovation, enhances user experiences, and solves problems that are too intricate for manual analysis. Machine learning is a subset of AI and mostly encompasses everything you have heard of in the new and otherwise.  

Outcome:

You can proficiently run ML algorithms in Jupiter notebooks locally and in a cloud provider for supervised, unsupervised and reinforcement learning. You can substitute in/out sections of the algorithms for better performance and can efficiently augment data and tune hyperparameters. You understand complex algorithms, model evaluation, overfitting, and bias-variance tradeoffs and know which models should be applied to different scenarios. Additionally, you have an in-depth understanding of XGBoost and can utilize top libraries for Python, including Scikit-learn, TensorFlow, Pandas, NumPy, and up-and-coming libraries such as fastai.

Key points:

• 6 hours a week for 2 weeks, including 10 labs
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Tuesday, January 7, 2024

Module 5

Deep Learning for PMs

Deep learning is a subset of machine learning where artificial neural networks, inspired by the human brain, learn from large amounts of data. It's particularly effective for complex tasks like image and speech recognition, natural language processing, and pattern detection, automatically identifying and extracting features without explicit programming.

Outcome:

You understand self-driving car algorithms (proximal policy optimization & Soft Actor Critic), where you will race your own virtual self-driving car against others in your cohort using the AWS DeepRacer platform. You can also explain neural networks, architectures (UNet, ResNet, Inception models and others), and frameworks well enough to be confident to present on this material. You have become proficient at running these algorithms both locally and in the cloud. You also have a solid understanding of Transformers and how muti-head attention works. 

Key points:

• 6 hours a week for 3 weeks, including 15 labs and 2 communications assignments
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Tuesday, January 24, 2024

Module 6

Generative AI for PMs

Generative Adversarial Networks (GANs), Diffusion Models, and Variational Autoencoders (VAEs) are pivotal in the field of generative modelling. GANs excel in generating high-quality, realistic data by training two neural networks competitively, making them invaluable for tasks like image synthesis. Diffusion models such as DALL-E and Sora are able to generate images or videos based on their training data. VAEs, known for their efficiency in learning complex data distributions, are crucial in unsupervised learning, enabling effective data compression and generation. Both technologies are instrumental in advancing areas such as image processing, data augmentation, and anomaly detection.

Outcome:

You understand Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and other models. You also have deep knowledge of Transformers and multi-head attention to the point you are comfortable presenting on this topic. 

Key points:

• 6 hours a week for 2 weeks, including 20 labs and 2 communications assignments
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Tuesday, February 11, 2024

Module 7

Generative AI & LLMs for PMs

Large Language Models (LLMs) are pivotal in the field of AI for their ability to process and generate human-like text. They excel in understanding context, nuance, and complexities of natural language, making them invaluable for tasks like translation (one spoken language to another, or a spoken language to code such as Python), content creation, customer service, personal assistants, market research, educational tools, healthcare, legal services, finance, and e-commerce. Additionally, they're becoming invaluable tools for developers, enhancing productivity and creativity in software development. LLMs can generate code snippets, translate code between languages, and even develop entire applications, transforming how programming is approached and executed in the tech industry.

Outcome:

You will deeply understand the available LLMs, including GPT-5, Bert Bloom, PaLM, and LLaMa. You will understand how to select and train one so you don't break the bank. You will also be able to recite the paper "Attention is all you need" and eloquently describe the types of transformers. You can improve your LLM responses with RAG, or fine-tuning using Parameter Efficient Fine Tuning (PEFT) with LoRA, adapters, and soft prompts. Using the latest algorithms and architectures, you can architect and deploy reinforcement learning from human feedback (RLHF) and minimize the human component.

Key points:

• 6 hours a week for 3 weeks, including 30 labs and 2 communications assignments
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Thursday, March 6, 2025

Module 8

Generative AI Agents for PMs

Generative AI agents are valuable for their ability to create new content like text, images, and code from scratch or by remixing training data in novel ways. This enables applications in automated content creation, natural language interaction, data augmentation, and further customizations for specific tasks. Additionally, it is possible to create AI Agents that outperform any individual LLM by creatively utilizing two or more LLMs together. 

Outcome:

You can stitch together any component LangSmith, LangChain, or Google's Vertex AI offer and create sophisticated architecture, including, but not limited to, using an LLM to generate summaries and themes from multiple text sources in PDF or machine-ingestible formats. You can create modern LLM-powered Agents for planning (task decomposition, self-reflection), memory (sensory, short-term memory, long-term memory), Maximum Inner Product Search (MIPS), and Tools (MRKL, TALM & Toolformer (finetuning for APIs), HuggingGPT)

Key points:

• 6 hours a week for 4 weeks, including 50 labs and 1 communications assignment
• Evaluation criteria: Class participation, lab(s), communication assignment(s), and a test.
• Next start date: Thursday, March 27, 2025

Module 15

0-1 Product Launch & Demo Day

There are two components to this capstone:

1. a group project where you will complete an AI/ML product end-to-end, from obtaining data and creating a database to leveraging data for an LLM and creating a front-end with Streamlit. To better understand what it is like working with a data science team, your 2 PM team will also get access to data science, data engineering, MLOps, full-stack dev, and a UX designer for a limited time.* You can think of it like a hackathon where you participate in both the PM and the technical elements.
2. Demo day. Here, you will showcase your projects to instructors, guest speakers, and external invitees, including VCs and senior product leaders. 

Outcome:

You can collaborate effectively with a team on an AI/ML product as you will be partnered with a data scientist, ML Engineer, data engineer, and full stack developer. Additionally, you can complete an AI/ML product end-to-end from obtaining data and creating a database to leveraging data for an LLM and creating a front-end with Streamlit all by yourselves with the help of an LLM, LangChain, and advice of other teammates. 

Key points:

• 9 hours a week for 7 weeks
• Evaluation criteria: Class participation, individual project submission, and group project submission.
• *One team may have three people if there is an odd number enrolled.
• Next start date: Tuesday, August 5, 2025

Module 16

Preparation and Final Exams for Artificial Intelligence Product Leader Certificate (AIPLC)^TM 

This will test your knowledge to its limits and could be compared to a few challenging university graduate-level exams plus a PhD comprehensive exam rolled into two days. 

Outcome:

Upon successfully passing the final exams, you will be rewarded with the certificate "Artificial Intelligence Product Leader", the most difficult-to-obtain credential in the world of AI/ML product management, placing you in the top echelon of AI/ML PMs. 

Key points:

• 1 Week of preparation, including mock tests, will be provided before the tests commence.
• Tests Day 1: Candidates complete a multiple choice exam, a short-form written exam, an essay, and a live lab. At the end of the day, candidates will also receive a take-home open-book question. Exams will take 5-8 hours.
• Tests Day 7: Candidates submit their responses to the open-book questions.
• Tests Day 10: Candidates take the equivalent of a PhD comprehensive exam with three instructors and a university professor who supervises PhD students. Candidates will also have to travel to the closest location in their city or the next closest city to be proctored in person for this.
• Students who pass the final exams will receive both a digital and a paper "Artificial Intelligence Product Leader" certificate handcrafted with significant attention to detail for you to commemorate this great accomplishment.
• To qualify for the final exams, you will need to meet the following requirements
  
  • To have demonstrated thought leadership with a minimum of 15 blogs (covered throughout the modules),
  • To have demonstrated your ability to complete an end-to-end AI/ML project 100% on your own (supported in Module 15)
  • If you are challenging the exam without first taking any of the Modules, you will still need to meet the prerequisites listed above and also need to pass a pre-qualifying exam, which consists of a Live Lab, verbal questions, and a written portion. Additionally, only those who, in an initial discussion with us, have proven that they have the potential to pass the pre-qualifying exam will be allowed to take it. We anticipate this number to be very small. 
• Next start date: Thursday, September 25, 2025