Research & Applied Projects
An overview of formal research endeavors, architectural implementations, and practical methodologies developed to handle data and reasoning constraints.
Transformer-Based Automated Subjective Answer Evaluation
Developed an evaluation pipeline utilizing BERT and SBERT alongside OCR to semantically score free-form subjective answers. The system normalizes language structure and establishes metric baselines against reviewer ground-truth.
Hybrid RNN-Transformer Efficiency Optimization
Architectural research investigating permutations of recurrent contexts with localized attention mechanisms to achieve standard language modeling capabilities under reduced computational budgets.
Mental Health Detection using Baseline NLP
A multi-label categorization architecture engineered to predict clinical mental health indications from unstructured text corpora utilizing attention-pooled BERT representations.
Autonomous Vehicle Driving Safety System
Orchestrated a vision pipeline with YOLOv8 for continuous inference of obstacle velocity, lane deviation, and collision probabilities in real time to support automated alerting functionality.
AI Tool for Data & Web Insights
LLM-integrated extraction architecture facilitating unstructured data ingestion, RAG-assisted contextual search, and formal structured output formatting.
Face Recognition Pipeline
Implemented localized CNN verification networks relying on FaceNet embeddings to achieve identity resolution against arbitrary topological variations.