Hi, I am Soumit Addanki, a Computer Science student at the University of Michigan - Ann Arbor. I am very passionate on everything computer science related, especially AI/ML. I aspire to leverage technology to create meaningful solutions that make a positive impact on the world.
RESUME.jpg)
My journey into software engineering started unexpectedly with AI—when I first encountered it, I was captivated by the potential of machine learning and data analytics. What began as curiosity quickly evolved into a passion, and since then, I've immersed myself in projects like building predictive models for the NFL Data Bowl and collaborating with the University of Michigan Men's Basketball team (See experiences for more).
Outside of the classroom, I enjoy hanging out with my friends, playing video games, and watching sports. Whether it's Poker Night or Sunday Night Football, I enjoy the thrill of strategy, the camaraderie that comes with competition, and the opportunity to sharpen my analytical skills while having fun.
My favorite video game, Elden Ring
The Squad
As Associate Manager at Stuart's of Novi, I led a team of 5+ employees, handled financial business, and analyzed sales data to optimize inventory and product offerings, improving workplace efficiency by 33% and lowering costs by 25%
Utilized Python (Pandas, NumPy) and R (dplyr, ggplot2) to analyze Catapult tracking data for the University of Michigan Men's Basketball team, boosting the players' training efficiency by 50%. Developed and deployed scalable machine learning models in Python (TensorFlow, scikit-learn, Pandas) to analyze over 1 million data points related to player performance and team dynamics. Implemented predictive analytics using advanced sports metrics to generate actionable insights into game strategies, improving model accuracy by 15%
Engineered a fully functional ray tracer in C++ which handled scenes with up to 1 million rays per frame and dropped rendering time by up to 50% for complex scenes through special partition optimization. Boosted the rendering pipeline using Monte Carlo integration, resulting in a 30% increase in rendering accuracy for light behavior simulations. Leveraged multi-threading and optimized memory management to accelerate rendering performance, reducing computation times by 40% in high-complexity environments.
Constructed an NLP-driven Social Media Post Summarizer leveraging techniques such as tokenization, stop-word removal, and TF-IDF vectorization, to condense posts school discussion platform into accurate summaries. Integrated Machine Learning algorithms such as transformer-based models (BERT, GPT) for supervised learning on large-scale datasets, achieving an amortized accuracy rate of 80%. Executed data preprocessing pipelines and model optimization to increase performance by 30% and reduce inference time by 20%.
Built a complete assembler in C that translates assembly code into machine code and links object files into executable programs. Designed and deployed a cache simulator, optimizing data access and reducing memory stall cycles by up to 20%. This was done in LC-2K, which follows a Von-Neumann and RISC architecture.