CV

B.E in Computer Engineering

• Coursework:
  • Machine Learning, Natural Language Processing, Big Data Analytics, Quantitative Analysis, Discrete Mathematics, Data Structures and Algorithms and Software Engineering.

Technical Diploma in Computer Engineering

• Coursework:
  • Fundamentals of Programming (C++, Python), Database Management System, Operating System, Computer Networks

Data Analyst Intern

• Engineered data preprocessing pipeline for 8+ datasets using Python and SQL, reducing data inconsistencies by 90% and standardizing analysis-ready datasets.
• Performed in-depth exploratory data analysis to identify key trends and patterns that lead to actionable business insights.
• Recreated World Bank data visualization, following Sir Hans Rosling’s style, showcasing data storytelling expertise.
• Developed an interactive Air Quality Analysis dashboard using Tableau, incorporating dynamic filters and diverse visualizations for comprehensive data storytelling.

Intracranial Aneurysm Detection

• Built a machine learning pipeline to automate the early detection and anatomical localization of brain aneurysms in multimodal 3D imaging (CTA, MRA, T1/T2 MRI) to prevent ruptures.
• Leveraged RSNA 2025 Kaggle Competition dataset to train the EffficientNet model on 4348 unique brain scans of patients.
• Preprocessed and cached 3D DICOM brain scans as 2D multi-channel tensors, reducing per-epoch processing time by 60x.
• Optimized training with Weighted Binary Cross Entropy achieving a validation loss of 0.25 for multi-label classification.

FinetuneX

• Built FinetuneX from scratch, a modular framework for fine-tuning LLMs with a Self-Implemented Transformer (decoder-only) architecture, implementing custom multi-head attention, positional encoding, and normalization layers based on model technical reports.
• Implemented supervised fine-tuning (SFT) with current support for Qwen2.5, including optimized training techniques for faster convergence.
• Developed a user-facing platform enabling dataset upload, training configuration selection, and base vs. fine-tuned model comparison.
• Designed the framework for extensibility, with planned support for additional architectures and post-training methods (LoRA, QLoRA, instruction tuning with human preference alignment).

FasalGuru

• Engineered an AI-powered agricultural assistant that delivers pest detection, irrigation guidance, fertilizer recommendations, and disease prevention tailored to real-time weather conditions.
• Fine-tuned and deployed a ResNet18 model for classifying 22 crop diseases, integrated with Ollama-hosted LLMs to generate dynamic prevention plans using live weather data.
• Trained Machine Learning models with Scikit-learn for irrigation and fertilizer prediction, enhancing precision farming capabilities.
• Enabled multilingual support across 200+ languages, expanding accessibility for diverse farming communities

• Led team to third place in MegaHack, a national-level hackathon at SJCEM, achieving a podium finish among 80+ teams.

• Secured 1st place among 100+ participants in Technical Quiz, demonstrating strong problem-solving and technical expertise.
• Earned Runner-Up in Code-a-thon, a competitive programming contest with 100+ participants, showcasing coding proficiency.

• Finalist in Technothon, an intercollege hackathon, recognized for technical innovation and effective collaboration under tight deadlines.

• As a Core Technical Member of Codecell at Rizvi College of Engineering, actively collaborated with the team to coordinate and enhance technical events, workshops, and hackathons, fostering a technology-driven culture and driving innovation across the campus.