SHOURYA
PUNJ_

A custom wedding web experience for Shourya & Ratna's wedding. Guests can watch the countdown tick to zero, share photos and videos from the celebration, and relive the memories together — all wrapped in a warm, game-inspired aesthetic.

• Wedding Countdown Timer
• Daily wordle & trivia with individual score tracking and team groom and team bride aggregations
• Share photos/videos as memories
• Wishes wall where family members send in their blessing

End-to-end image recognition pipeline built using two services. Service one reads the input images from an input s3 bucket and sends them to a queue. Service two reads images from the queue and uses AWS Rekognition model to identify objects in the images and store them in output bucket.

• Classifying objects in the image
• Hands on learning experience with AWS services.

ML model using Apache Spark’s MLlib libraries and trained it over AWS EMR in parallel consisting of 1 master and 3 workers. Saved the model in s3 and then used to predict the wine quality on test data. Created a container using docker for the ML model to simplify deployment and pushed the container to docker hub. ML training and prediction code was developed in Python

• Prediction of wine quality
• Exposure to AWS services as Master's project

Daily crawler scraping tender postings across hundreds of vendor and government procurement portals. A deduplication and normalisation layer feeds a sales-enablement tool that surfaces relevant opportunities.

• Hundreds of tenders ingested per day with high fidelity.
• Cut manual sourcing effort by an order of magnitude.
• Built scraper-resilience tooling to handle frequent site changes.