My Builds.

Developed a spatio-temporal traffic forecasting system combining classical baselines (Naive, ARIMA, XGBoost), GRU-based deep learning, and Temporal Fusion Transformer (TFT) for state-of-the-art probabilistic predictions. Implemented MC Dropout and quantile regression for uncertainty quantification and cross-junction dependency modeling. Designed a modular Python codebase with exploratory data analysis and visualization achieving robust RMSE(5.9849) and MAE(4.0017) performance and actionable uncertainty insights for urban traffic planning.

Developed an AI-powered BIM quality assurance platform combining deterministic IFC rule validation and LLM-based analysis using Gemini/OpenAI/MiniMax. Integrated buildingSMART Data Dictionary (bSDD) for standardized property checks, terminology normalization, and multi-language support. Implemented readiness scoring, auto-fix capabilities, and detailed reporting for IFC models, reducing property and naming inconsistencies by ~40%. Designed a modular FastAPI + Next.js architecture enabling scalable validation pipelines, interactive dashboards, and exportable compliance reports.

IIITDM

Aug 2023

An Institute of National Importance Website, where I re-innovated on completed frontend and some of old backend of the website, created a custom CMS for the admin to manage the content of the website, resulting in a remarkable 95% performance improvement in rendering, and achieving a flawless 100% SEO and best practice score

Operator I

Mar 2023

Designed and implemented an API and mobile application for an academic project focused on advanced image processing techniques. Integrated arithmetic operations, spatial domain transformations, and frequency domain filtering using convolution-based techniques.

This industry project focused on developing an image processing pipeline to detect and analyze gas burner holes using OpenCV and Hough Transform. It enhanced visibility through contrast stretching, applied segmentation for accurate detection, and extracted key measurements like hole count and size distribution. Masking and filtering refined the analysis, while automated CSV reports enabled efficient data evaluation.

Carnet OS

April 2022

A database management system exploring project on ticketing system of railways, Innovated UI into a macOS clone with subtle animations using popmotion and winbox, learnt about using various ORMs like Prisma, Drizzle, etc., along with direct SQL query execution.

Developed a machine learning system for emotion recognition using ECG signals, extracting time-series features like HRV metrics (RMSSD, SDNN, NNx, CSI, CVI). Implemented and evaluated Decision Tree and Random Forest classifiers to predict emotions such as happiness, anger, and disgust.

Gossip Box

November 2021

A mobile social media chat application which includes features inspired from snapchat and whatsapp, where I learned about real time streaming data to update chat, worked with lottie files for animation and firebase for backend.

Developed facial recognition and traffic sign recognition systems using machine learning and deep learning techniques. For facial recognition, implemented Principal Component Analysis (PCA) for dimensionality reduction and used KNN, LDA, and Logistic Regression models. For traffic sign recognition, applied CNNs and utilized the GTRSB dataset for training. Leveraged OpenCV and PIL for image processing, gaining hands-on experience with real-world data collection and deep learning concepts.