Four years,in order.

Specialisation in Artificial Intelligence and Machine Learning.

• Coursework across machine learning, deep learning, computer vision and explainable AI, carried into project work rather than left on the transcript.
• Most of the systems in this portfolio started as coursework and kept going after the deadline.

Led a 40+ member visual media team.

• Owned the club’s visual output end to end — direction, shoot planning, edit and delivery.
• Ran a 40+ member team across recurring shows and campus-wide campaigns.

Weld inspection was manual and inconsistent between inspectors. I owned an automated defect-localization pipeline end to end.

• Fine-tuned a YOLOv8-L detector on high-resolution weld imagery, owning the pipeline from data to evaluation.
• Curated and labelled the dataset, ran 100 GPU training epochs, and evaluated on precision, recall, F1 and mAP — 99.1% on training against 74.3% held-out.
• Diagnosed the gap as overfitting, revised the dataset and splits, and presented the trade-offs and next steps directly to the scientific staff.

• YOLOv8
• PyTorch
• Ultralytics
• OpenCV
• Python

Led a 25+ member team across sponsorship, publicity and media for two flagship events.

• Coordinated outreach for VIT Chennai’s technical and cultural flagships in the same cycle.
• Ran sponsorship, publicity and media as one function rather than three disconnected teams.

Presented at the 2nd International Conference on Advanced Nexus of Data and Information Technology.

• Event-driven Boto3 system scanning EC2, EBS, S3, IAM and RDS for waste, reclaiming it on a scheduled Lambda cron.
• Dry-run simulation and protected-tag exclusions guard against accidental deletes.

• Python
• Boto3
• AWS Lambda
• scikit-learn