TECHNICAL SKILLS

• Programming Languages: Python, C/C++, SQL
• Machine Learning/AI: Pytorch, Hugging Face Transformers, NLTK, SpaCy, Sk-learn, XGBoost, OpenCV
• Cloud Platforms: GCP, AWS
• Programming Skills: Algorithms and Data Structures, Trees, Graphs, Git, SQL, Docker

PROJECTS

Radiographic Bone Fracture Classification in X-ray Images

• Created a bone fracture classification system using deep learning on X-ray images, leveraging transfer learning with Hugging Face pretrained models, fine-tuned on the MURA dataset, a large collection of 40,000+ musculoskeletal radiographs across seven anatomical regions.
• Utilized CycleGAN and StyleGAN to generate synthetic X-ray images, addressing dataset limitations and enhancing model generalization. Implemented various neural network architectures (U-Net, Attention U-Net, ResNet, DenseNet, Vision Transformers, Hybrid ViT) using PyTorch.
• Achieved optimal performance with DenseNet: 97.2% accuracy, 0.97 precision, 0.96 sensitivity, 0.95 F1 score, and 0.98 AUC-ROC, demonstrating the system’s robustness and reliability in medical imaging tasks.

GNN-based Recommender System for Movie Recommendations

• Built a GNN-based recommender system to predict movie ratings on the MovieLens 100K dataset by modeling interactions between users and movies as a bipartite graph, leveraging message passing and graph convolutions.
• Processed raw data, constructed user and movie node features, created graph edges, and incorporated collaborative filtering techniques.
• Trained using PyTorch Geometric, achieved optimal results with RMSE of 0.90, MAE of 0.72, and precision@K of 0.85.

LLM Jailbreaking

• Explored security vulnerabilities in large language models, identifying weaknesses that could be exploited for unintended behaviors.
• Designed a framework to test, document, and categorize jailbreak techniques, evaluating their effectiveness and reproducibility.
• Provided recommendations for robust model defenses focusing on AI security, bias mitigation, and ethical deployment.

WORK EXPERIENCE

Machine Learning Engineer Intern, Treevah LLC (Feb 2025 - Present)

• Designed a file classification system achieving 98.5% accuracy using NLP models, processing 2.3M+ files across 27 categories with 400ms average inference time, reducing manual labeling efforts by 73%.
• Integrated classification system with AWS infrastructure (S3, Lambda, EC2), achieving 99.97% uptime SLA while handling 1,250 files/sec at scale, reducing cloud processing costs by 40% through auto-scaling optimization.

Data Science Engineer, Hack For LA (Feb 2025 - Present)

• Integrated and cleaned 5+ years of LA Metro transit data (2019-2025) from GTFS schedules, real-time APIs, and external datasets, enabling analysis of 25.3M+ monthly boardings and recovery to 82.9% of pre-pandemic ridership levels.
• Performed EDA on 300M+ annual boardings, identifying 5.6% YoY bus ridership growth post bus-lane enforcement implementation and correlations between $293 fines for lane violations and 8% weekend service reliability improvements.

EDUCATION

M.S. in Computer Engineering, New York University (Sept 2022 – Dec 2024)
Relevant Coursework: Data Structures, Deep Learning, Reinforcement Learning, Advanced ML, ML for Cybersecurity, Probability & Stochastic Processes
GPA: 3.67/4.0

B.Tech in Electronics and Communication Engineering, GGSIPU Delhi, India (Aug 2017 – July 2021)