Hello! I am Albert, a mentor, researcher, and Computer Science student at the University of Texas at Austin.
As a mentor and developer, I lead by example in being patient and persistent while approaching opportunities with unique and open perspectives.
For example, I am a computer science intructor for Juni Learning, where I teach important computer science concepts in Java, C++, Python, or Scratch to the next generation of students. I also perform Autonomous Robotics
and Learning Agents research at the University of Texas at Austin, where I hold responsiblities in developing specialized learning applications to analyze human-robot interaction. Academically, I commit to my education and excel in my studies with a GPA of 3.96.
In addition to school and work, I constantly seek opportunities to expand my knowledge and am interested in Software Engineering, Machine Learning, and Data Science applications.
Human-dialogue systems are useful for the execution of a wide variety of human-robot scenarios.
These systems are able to continuously learn previously experienced language concepts and utilize them
in new environments without requiring its human clients to use machine-understandable languages.
For example, if a person requests "bring Bob a white mug", the
system must recognize that Bob is a person represented as a destination and that "white" and "mug" are properties
that describe an object in the environment. The objective was to produce a high performance C++ semantic
parser with memory usage reductions and reduced element access times for future time-sensitive scenarios and real time parsing.
Skills: C++, Python
Research Project's GitHub
Researchers constantly move
in and out of labs with access to expensive and private equipment. Without
proper security, anyone would be able access this equipment and cause a major security hazard! To combat this issue, we programmed the webcams on
the lab's laptops and BWI Robots to accomplish facial recognition — a powerful tool for authentication. Creating
facial recognition with standard neural networks require incredibly large amounts of training data, so we utilized
one-shot learning using siamese networks. This parallel network allowed us to store and verify users in under a few seconds with minimal data (30+ frames)! The applications of this type of facial recognition also extend past labs to improve the security of users with camera devices.
Skills: Python, Siamese One-Shot Learning, OpenCV
Research Project's GitHub
Customers who shop at large-scale clothing stores are often faced with an annoying problem — high volumes of traffic in fitting rooms. As part of the Retail Tech build team,
we decided to tackle this issue through automation. Through an app, customers can notify the system that they desire a fitting room. By doing so, they can keep browsing while not having to wait in line!
When a room is ready, the customer will be notified by text or email to enter the room. Once finished, they can simply use the tablet device outside to sign out.
Now, why is this automation useful? First, it increases a business's resources by getting rid of the need for a designated fitting room employee. Second, it saves the customer time. Third, it increases
store interaction and boosts the number of returning customers because of the lack of a line. What does all this mean? Profit.
Skills: Java, Android Studio
Build Team Project's GitHub
Texas Convergent Website
Every game has flaws in its coding and security which is why many developers continuously release updates to fix them to improve their playerbase's experience.
To find these flaws and understand how malicious players capitalize on them, we, CSAIM, aim to reverse engineer functions of Counter-Strike that grant them
unfair mechanical and visual advantages. In order to do so, we acquired offsets from local files and memory to debug and recreate recorded problems. Using a mix of
linear algebra for mapping and Windows memory functions, we were able to develop the real-time software that grants these advantages. In the end, its source code
allows the game's developers to create new methods to combat these recurring problems that ruin other players' experiences. However, publishing these online
allows individuals to find and use these maliciously. For this reason, they are often sent directly to developers.
Skills: C++, Linear Algebra
Released Open Source Version GitHub
My friends and I were always faced with the grueling task of deciding where to eat. Rather than using Google Maps to retrieve an endless list of nearby restaurants, we worked to create
a restaurant recommending app that factors in user input preferences, distance from current location, and reviews from Zomato. It is able to select the top five open restaurants based on
weighted score using our formula. After learning about machine learning from Stanford Online, I hope to be able to utilize text flagging/classification on reviews for more accurate and user-
Skills: Python, Android Studio, Google Places + Zomato API
After fiddling around with the Adobe Suite for a while, I thought to myself, why not start a cheap graphic design service with my current skills? I found my first set of clients
through a gaming community called Steam. They were incredibly satisfied with my logos for the price and referred their friends to me. Over the next year, this cycle continued, and I found myself improving
and picking up new skills such as animation and photo editing. I created a simple Google Sites website for clients to contact me through, and customers from outside the gaming community started coming in. This included random people from the internet
and companies such as Hanoi TechLink and GenXComm. Since 2015, I have had over 200 satisfied clients and have completed 250+ projects — some of which were free! I have also created a new website for the business.
Skills: Adobe Photoshop, Adobe Premiere Pro, Adobe After Effects
Old Tempest Designs Website
New Tempest Designs Website