Hi, I am Aryan Dua, a passionate and inquisitive 4th year computer science student at IIT Delhi, on a journey to unravel the intricacies of the digital realm. My academic voyage has been a fusion of rigorous learning and hands-on exploration, equipping me with a diverse skill set ranging from algorithmic problem-solving to software development. Curiosity fuels my drive, pushing me to delve into emerging technologies and their potential to reshape our world. With an innate love for elegant code and a knack for collaborative teamwork, I relish tackling complex challenges and translating ideas into innovative solutions. Beyond the screen, I enjoy leveraging my skills to contribute positively to my community, making technology accessible and beneficial for all. Join me as I continue to unravel the limitless possibilities that computer science offers.
Here, check out some of my interests:
Data Structs and Algorithms
Artificial Intelligence
Machine Learning
Advanced AI
NLP
Microeconomics
Some techincal skills I have learnt along my journey are:
I am in my 9th semester of my degree at IIT-Delhi. I have been in the merit list of the "top 7% students in each semester" two times so far.
I cleared the prestigious JEE Advanced entrance examination in 2020 with an All India Rank of 516, out of the 250,000 students who gave it.
I cleared the JEE Mains entrance examination in 2020 with an All India Rank of 534, out of the 1,000,000 students who gave it.
I cleared the prestigious KVPY scholarship examination in 2020 with an All India Rank of 677, and was awarded the scholarship. I also scored 418/450 in the BITSAT entrance examination that tests English, science, maths and logic. I was in the top 100 students among those who appeared for it.
I was one of the few who cleared the NSEC(National Standard Examination of Chemistry) and was qualified for the Indian National Chemistry Olympiad 2020. The qualification rate in the exam was 1%. I had also cleared the regional round of the Maths Olympiad with a state rank of 12.
Programming asynchronous algorithms which involve gradients, projections and proximity operators and determining hyperparameter selection strategies for the synchronous and asynchronous version of the algorithms and comparing them both.
Successfully experimented with the implementation of the PROGPROMPT paper which presented a programmatic LLM prompt structure that enables plan generation functional across situated environments, robot capabilities, and tasks. The key insight was to prompt the LLM with program-like specifications of the available actions and objects in an environment, as well as with example programs that can be executed.
Implemented a mini-batching streaming system for counting words in a stream of tweets and stored the counts to Redis. Ensured Redis fault tolerance by checkpointing, and worker fault tolerance by updating Redis atomically using Lua.
From open source data, I built a web application using Django with a Postgres-12 server at the backend. The application lets you choose the filters to apply for certain statistics of a country, displays the list of countries that satisfy the filter and plots Covid-19 graphs for those countries.
Attempted the hard track (vs easy track) assignments based on the latest Linux Kernel. Implemented everything from scheduling to memory to device drivers in the kernel. Was privileged to be a part of the first Operating Systems course in the world to teach concepts practically through the Linux Kernel. We spent 4 months looking at the intricacies of the kernel code during lectures and applied these concepts in assignments wherein we had to look through the kernel documentation and elixir.bootlin. The most challenging aspect of this project was to implement OS task scheduling. We implemented a thread-safe data structure in the Linux kernel to track the number of context switches of processes. We added system calls to the kernel for registering and de-registering processes and returning the number of switches.
Implemented a general neural network architecture to learn a model for multi-class classification, working on Fashion- MNIST data set (a data set of Zalando's article images, where each example is represented as a 28 x 28 grayscale image, with a label from 10 classes)
Applied search-based methods for determining possible corrections for the input erroneous text.
Implemented and constructed a text classifier using the Naive Bayes Algorithm which works on a Large movie review data set and predicts the sentiment of a movie review as positive or negative. Also formed a word cloud of the most frequently occurring words of the positive and negative classes.
Designed a scalable and efficient multithreaded Server-Client Arch. for the distribution of a decentralized file to users.
A muliplayer maze game based on the IIT Delhi map, built using SDL2 library in C++. The multiplayer dseign has been implemented using sockets. The game is a basic life simulator of life at IITD. You have 4 coefficients that determine your "score", and be sure not to collide into the angry professor. Instructions to download and play are there in README.
Built a multi-label SVM classifier which can classify the type of actuator required based on the input features like Stress, Strain, Efficiency, etc. The data collected from the research site was sparse and incomplete and so I had to implement a novel approach to build the classifier with an accuracy of 72% as of now. Instead of using all 5 feature columns to build the classifer, I had to use them 2 at a time to maximise training examples with all non-null features and then build 10 classifiers. I then found the final prediction by compiling the predictions from all the classifiers.
Developed a fully operational processor using VHDL, capable of executing a comprehensive range of machine-level instructions. Upon translating assembly programs into their corresponding binary representations, the processor reliably performs the specified sequence of commands.
Built a full compiler which first converts any given program written in the WHILE programming language to an Abstract Syntax Tree, then from that AST it evaluates the program using a VMC Stack Machine implementation(Value - Memory - Command). The code of the compiler was written in sml(ML-LEX and ML-YACC were used for lexing and parsing the program)
Developed a fundamental audio processing library capable of detecting a specified set of twelve predetermined keywords from a one-second audio segment. The machine learning algorithm employed for this task was crafted using C++. However, the project encompassed multiple programming languages including C++, Python, bash scripting, and Makefiles.
Implemented the merge sort algorithm in Assembly language, sorting sets of strings given in input files, lexicographically. This was simulated using ARMSim.
Built a program to detect edges in an image using the properties of high and low-pass filters, 2D convolutions, Gaussian kernels, Fourier transforms, etc. The novel part about this project was that we did not use the built-in OpenCV library functions, we implemented the whole program ourselves, from scratch.
Developing and implementing my own Cryptocurrency, as well as the software for processing transactions through data structures in Java.
Developing machine learning models to solve problems based on classification, neural networks, supervised and unsupervised learning algorithms.
"I am thrilled to offer my wholehearted recommendation for Aryan. Throughout his time with us, he consistently showcased strong analytical abilities, a keen grasp of complex concepts, and a genuine enthusiasm for learning. Witnessing his exceptional growth and unwavering dedication as a data scientist was a privilege. Aryan's relentless pursuit of excellence and positive attitude left a significant impact on our team, and I am confident that he will flourish in his data science career."
"Aryan is an excellent programmer and researcher -- I was repeatedly impressed by how quickly he grasped onto graduate-level topics in mathematics, as well as how he was very self-driven in both learning theory and developing code."
"Aryan's determination to success led him to excel in class 12th. One thing which differentiates Aryan from other students is his curiosity and the habit of asking questions. He is very organised, consistent and persistent. Aryan has a growth mindset and able to connect learning to life. He always sets his goal and also knows how to deal with failure."
"Aryan was an outstanding student. He was always a self-learner. Prior to each class, he would read and try to understand the material by himself and ask doubts to deepen his understanding. From my understanding, Aryan is a sincere learner, affectionate towards his friends and teachers and hardworking individual."