Computer Science Masters

As of 12-01-2026 I have taken 8/10 classes for the OMSCS program.

KBAI - Knowledge Based Artificial Intelligence

Core Themes

BFS - Breadth First Search
DFS - Depth First Search
A* Search
Heuristics

The semester long, core project of the class is designing a software agent to solve the Raven’s Progressive Matrices problems. A series of visual tests that measures human’s intelligence. An example is below.

💡

Takeaway: When designing software, it’s extremely relevant to understand the layered cognitive reasoning and problem solving steps humans go through; translating those human techniques into code.

🔥

Score: 8/10. I enjoyed the cognitive science approach to software, albeit the lectures were not engaging and repetitive. The mini-projects in KBAI were well designed, engaging and a lot of fun!

ML4T - Machine Learning for Trading

Core Themes

Software in the Stock Market
- Stock Market Financial Overview
- Technical Indicators
- High Frequency Trading (HFT)
Linear Regression
Decision Trees - Random Forest Learner
Reinforcement Learning - Q-Learner/Dyna

The final project in the class combines the prior machine learning techniques. The goal of the final project is to create a machine learning model where-in the rules are dynamically determined based on a trained supervised classification random forest model. The model was then deployed in a hypothetical stock simulation where decisions are made to buy and sell stock depending on the model’s attributes and the subsequent technical indicators.

The screenshot below depicts my trained model, the Strategy Learner, outperforming the portfolio of the Benchmark and Manual Strategy.

💡

Takeaway: Machine Learning can be fairly digestable to understand and implement, it’s a lot of math implemented in such a way that can surface amazing insights and predictability. Machine Learning can be deployed in a lot of industries and the analytical, productivity and revenue gains can be monumental.

🔥

Score: 10/10. I’m fascinated with parts of the finance industry. Dr. Balch is a great lecturer, the content and projects were very engaging.

The classic iris flower classification example is a great dataset to get your feet wet in Machine Learning!!

SDP - Software Development Process

Core Themes

Mobile Application Development (JAVA) - Android Studio
Git/GitHub
Blackbox Testing
Whitebox Testing
- Path, Statement, Branch Coverage
Unit Test Generation

The larger group project in this class is developing an application in JAVA, using Android Studio. The loose structure of working on software in a typical SDLC process was something I’m used to in my full time work. However, it was a great learning experience in the context of JAVA and Android Studio.

The whitebox testing assignments were very engaging and reinforced learning about the different coverage techniques.

The screenshot below is the entry panel of the application my team and I developed.

💡

Takeaway: Testing code is critical. Prior to this class I thought I was good at testing edge cases but I did so informally. Learning about whitebox testing helped formalize the quality assurance process that is often overlooked or underutilized in software development.

🔥

Score: 9/10 This was my first exposure to formal software testing, I found it fascinating to learn about the different coverage techniques for unit tests. Additionally the group project was a lot of fun. JAVA and mobile development is something I’ve experimented with but never worked a lot on. Getting back into JAVA helped my Python skills.

AI4R - Artificial Intelligence for Robotics

Core Themes

Histogram Filter
Kalman Filter
Particle Filter
Kinematic Bicycle Model
PID Controller (Proportional, Integral, Derivative)
SLAM (Simultaneous Localization and Mapping)
Search (A-star, Dynamic Programming (value iteration, optimum policy))

All of AI44’s projects are well thought-out and engaging. To complete the projects in full, you must understand the concepts at a fairly deep level. Each project comes with visual GUI interactions of the tasks, involving flying a drone to a certain height or localizing a moving asteroid.

The following gif is taken from the lectures, visualizing particle filters being constantly sampled and eventually through a particle filter, localizing on the target.

The following visual is a demonstration of the Kalman filter. Predictions and measurements are fed into a filter, based on weights of each model, the prediction and measurements overtime converges into the estimated position of the vehicle.

Kalman Filter - Part 1

Lastly, extra credit for the class was implementing a Kalman Filter within moving physical hardware. The following video demonstrates my imperfect Kalman Filter robot turning and stopping autonomously based on ultra sonic measurements.

https://www.youtube.com/watch?v=K2woOHnF7-0

💡

Takeaway: Localization is the core of robotics. Robots can’t perform tasks or actions if they don’t know where they are in the world. There’s a lot of concepts that go into mapping and localization for autonomous vehicles; these concepts are foundational to robotics, even in the age of AI.

🔥

Score: 8/10 I learned the most array of concepts in this class. The projects were approachable and the concepts are not terribly mind-bending. Through the projects and the lectures, the instructors give just enough to initially cling to the concepts and then further research is required on the student to fully grasp the ideas.

CN - Computer Networks

Core Themes

The Open Systems Interconnection (OSI) Model
- 7 Layer Framework
iBGP/eBGP
TCP/UDP
Congestion Control
PyBGPStream
BGP Hijacking/Measurement
Distance Vector Routing
SDN Firewall

OSI Model

💡

Takeaway: Computer Networks are extremely intricate. It’s fascinating the complexity that lives beneath mundane daily tasks for most of the world, checking your bank account, loading a YouTube video, refreshing a page. Given that, there is heavy abstraction and interoperability that has made networking much easier, such as SDN. However, the foundations of networking are extremely complex.

🔥

Score: 6.5/10 CN was a flurry of information. The lectures were tough to follow and grasp; I personally did not have much interest in the deeper concepts of networking. The projects however were fun and overall the course was fairly easy.

VGD - Video Game Design & Programming

Core Themes (Unity)

AI Tracking Agents
Animation Mecanim
Game Feel
Game Physics
Game Storytelling

The largest aspect of VGD is a group project to design and build a 3D game. Our group created a maze based, reward finding, combat game called Maze Runner. Below is our game trailer.

https://www.youtube.com/watch?v=fKRjrjPJjEE&feature=youtu.be

💡

Takeaway: Unity is incredibly powerful. I dare to say it’s ‘easy’ to make a 3D game in 2025. However, Unity provides all the tools and resources to be able to craft really modern games from your bedroom. There is an extremely steep learning curve and like anything else, there are multiple ways to accomplish a desired outcome.

🔥

Score: 5/10 The course is not too challenging. I enjoyed poking around in Unity and creating a game from scratch but I am personally not interested in the intricacies of game development. I enjoyed brushing up on C#. The lectures are well done if you have an interest in this field.

AIES - Artificial Intelligence Ethics & Society

Core Themes

Fairness, Bias, Legality of Data Collection
Fairness in AL/ML
Bias Mitigation in AI
- Disparate Impact
- Statistical Parity

AIES centered around bias mitigation, particularly around analyzing datasets to uncover fairness, or lack of. The following is a graph I built from a prison dataset, analyzing the percentage of custody statuses by Gender.

For my final paper, I analyzed a piece of research that compared LLM hiring decisions across different LLM models, namely they compared a proprietary in-house supervised trained model with other big generic LLMs like GPT and Claude.

💡

Takeaway: This course highlights important externalities that come with widespread internet usage, particularly in a time of booming Large Language Models that vacuum up every bit of data in it’s path. As developers, we are entrusted to design software that accurately reflect normative behavior and conversely mitigate non-normative behavior and bias.

🔥

Score: 7/10 I enjoyed this class, it is not challenging but does get your gears turning on the impact software and data design can have. Data is my professional niche; it’s immensely important negative externalities are constantly and closely examined.

NLP - Natural Language Processing

Core Themes

RNN
Transformers
Information Retrieval
Open Domain Question Answering
Machine Extraction
Machine Reading
Machine Translation
- Early Seq2Seq Models, Phrase Lattices and Transformers
Privacy & Societal Impacts
- Differential Privacy, Clipping Gradients

💡

Takeaway: NLP is an extremely relevant course in 2026 with the explosion of LLMs. This course was fascinating to understand how LLMs function at their core. I appreciated the structure of the course starting from earlier language model architectures that are no longer used but laid the ground for LLMs like GPT. You sometimes hear LLMs are just supercharged predictive machines, constantly predicting the next token, in this course you actually understand how those predictions are made. I don’t believe this course was surface level either, the coding had bits and pieces of building language models from scratch but also introducing popular NLP python libraries. The course really tested your conceptual understanding as well, as the exams and quizzes were extremely tough in this regard.

🔥

Score: 9/10 There are some modules that could be re-worked but Professor Riedl is an excellent teacher and most of the lectures are designed extremely well. I thought this was one of the best courses for learning. I did think some of the coding exercises were a little too hand-holding at the start of the semester but I ultimately think everything was rounded out nicely.

Remaining Classes through Fall 2026:

Computer Science Masters - Course Review