About Me

Daniel Mohanadhas Link to heading

Computer vision, natural language processing, cybersecurity, and their intersections.

The University of Texas at San Antonio (UTSA)
- PhD in Computer Science, GRA & TA (Computer and Information Security, Unix and Network Security)
- Expected Graduation: Dec 2027
- Advised by: Dr. Amanda Fernandez
The University of Texas at San Antonio (UTSA)
- B.S. in Computer Science, Concentrations in Cybersecurity & Data Science
- Graduation: May 2023
- GPA: 3.59

Computer Languages: Python, C/C++, Bash, Java
Libraries/Framework: PyTorch, TensorFlow, OpenCV
Software & Tools: Git/GitHub, Docker, LaTeX, Microsoft Office, VSCode, OpenAI API
Soft: Self-learner, teamwork, conflict resolution, critical thinking, project leadership/management

Graduate Research Assistant | Vision and Artificial Intelligence Lab | UTSA (July 2023 - Present)
- Investigated classification approaches on images affected by adverse conditions using few-shot and GANs
- Team lead over 4 people, responsible for mentoring 3 students, communication with client, project completion
- Used diffusion models to create a synthetic dataset of thousands of images to augment a source of 20 images
- Investigating semantic segmentation of images with SAM2 with multi-modal prompts
Research Assistant | INSURE Program | DoD (Jan 2023 - April 2023)
- Created an NLP model to detect malicious files in binary, assembly, or source code forms using opcodes
- Found and created data of interest, re-evaluated and re-defined scope throughout the project with client
- Designed an end-to-end pipeline from data collection to model training/testing; achieved 85% accuracy
Research Assistant | Vision and Artificial Intelligence Lab | UTSA (May 2022 - July 2022)
- Worked in a multi-disciplinary team to reduce the time it takes to analyze images of nuclear fuel
- Customized a PyTorch framework to segment images in seconds instead of 6+ hours. Poster presentation.

D. Mohanadhas et al, “Improving Classification Via Generative Techniques On Hypersonic Sensor Data" in U.S. Journal for Hypersonics, vol. 1, 2025 (Submitted)
D. Mohanadhas et al, “Multimodal Approaches for Realistic Synthetic Data Generation in Materials Fabrication," in CVPR Workshop on Multimodal Learning for Materials Science, 2025. [Accepted]
C. Snyder et al., “Semantic Segmentation of Micrographs for Nuclear Fuel Analysis and Degradation Quantification,” in IEEE Access, vol. 11, pp. 118512-118520, 2023, doi: 10.1109/ACCESS.2023.3326074

AI-Generated Detection Browser Extension (Oct 2024 - Dec 2024)
- Achieved a 95% test accuracy on the ELSA D3 dataset (2.5 TB) with an EfficientNet backbone
- Further finetuned model on numerous other datasets, but concluded my goal of creating a single model to detect all AI-generated images was unfeasible, as each AI generator model has new artifacts to detect. Website
Rowdy Datathon (Oct 2024)
- Only person to find a statistically significant correlation between the Journey North and USDA PDP datasets
- Web scraped to create a dataset of butterfly population by county across 30 years. Filtered outliers and artifacts
- Achieved an r^2 of .87 based on various machine learning models. Code and final datasets
Natural Language Processing (Feb 2024 – Mar 2024)
- Improved prompt engineering, implemented Retreival Augmented Generation (RAG) with google-T5
- Used OpenAI API to create a custom dataset to finetune Llama2, Llama3, Phi2, and Mistral
Computer and Information Security (Sep 2022 - Dec 2022)
- Used Hashcat and John the Ripper to crack 20 passwords using custom rules and character sets
- Attacked 10 Intel and ARM machines using buffer overflow, ret2libc, string format, and other exploits
- Cracked 12 wireless networks using WEP and WPA encryption with arpreplay and chopchop attacks

Honor Roll | UTSA (Jan 2022 - Present)
Texas A&M Rules and Regulations Committee (Mar 2020 - Aug 2020)
Spotlight Paper | Multimodal Learning for Materials Science Workshop (June 2025)