About Me
Hey! I'm Hemansh, and I'm currently a Master of Engineering student at Virginia Tech, where I'm studying Computer Science with a focus on Data Analytics and AI. When I'm not working on projects, I'm usually reading, playing video games, at the gym, or hanging out with friends and family.
The Path to Now
My start in college was a bit of a reality check. During my 1st year at George Mason, I decided to take a TensorFlow course on Udemy after hearing the buzz around "AI". It was a steep learning curve and probably a bit too advanced for a beginner, and although I probably didn't fully grasp all of the concepts, it forced me to learn the fundamentals of building data science and model building.
I decided to transfer to Virginia Tech my 2nd year to study Computational Modeling and Data Analytics because of my interest in data and AI. This proved to be a great decision as the amount of opportunities I got here were tremendous. I made a lot of new friends and also spent two years as a researcher at the Hume Center, where I got my first taste of working on technical teams using GitLab and training a custom-built Reinforcement Learning agent from scratch using auto-generated testing data. But the whole process — from designing to testing to the constant dimensionality bugs in PyTorch — made me fall in love with this subject even more.
After graduating with my B.S. in May 2025, I transitioned into the Master's program here at VT. I've really enjoyed the depth of graduate-level coursework; it provides a level of intuition that goes far beyond the surface. My focus has now shifted toward making AI, dashboards, and analysis production-ready.
After spending my third year of undergrad formalizing my data science/ML knowledge, I'm now focused on using various cloud platforms to take my models and visualizations out of a notebook and put them into the world for people to actually use.
Education
George Mason University
Aug 2022 – May 2023B.S. Computer Science
Fairfax, VA
Virginia Tech
Aug 2023 – May 2025B.S. Computational Modeling & Data Analytics
Blacksburg, VA
Virginia Tech
Aug 2025 – Dec 2026M.Eng. Computer Science
Alexandria, VA · Expected Dec 2026
Experience
Hume Center for National Security and Technology
AI/ML Researcher
- Engineered an NLP-to-YAML data pipeline using a 70B parameter LLM (Llama-3.3 via Groq API) and a FastAPI backend; this tool parses natural language prompts to auto-generate valid network topologies (YAML files) for robust, scaled-environment RL training.
- Designed, trained, and benchmarked Reinforcement Learning agents (Tabular Q-learning, Deep Q-Network) in PyTorch to perform cyber-network vulnerability analysis within the NASim simulation environment.
- Optimized DQN agent sample efficiency and stability by implementing a target network and replay memory buffer; achieved a 20% increase in average reward on test networks, demonstrating a massive performance lift over the baseline NASim DQN-agent.
- Conducted comparative analysis across agents (Tabular Q-learning vs. DQN) and strategies (Epsilon-Greedy vs. UCB). Used paired t-tests to determine statistical significance of reward differences, and benchmarked all models by visualizing reward convergence, DQN loss curves, and bootstrapped 95% confidence intervals.
Featured Projects
A selection of recent work.
AI-powered crypto portfolio advisor combining neural collaborative filtering and real-time market regime analysis. Personalized token recommendations backed by 200k+ wallet interactions and dynamic risk optimization.
A comprehensive geospatial and statistical analysis of NYC Taxi trip data, combining backend statistical modeling with a production-ready interactive dashboard.
Stress and emotional detection using the WESAD dataset with consumer-grade and medical-grade wearable sensors to identify physiological stress markers.
Technical Skills
Languages
Frameworks
Tools
Contact
I'm always open to new opportunities, collaborations, or just a chat. Feel free to reach out!