Projects

Real-time AI, deep learning and data engineering work.

Built a role-based trekking management web application for the IIT Madras Diploma App Dev I project, supporting Admin, Trek Staff and Trekker workflows.
Implemented trek discovery, filtering, booking, slot reservations, staff assignment and booking approval workflows with role-aware access control.
Created modular Flask routes, SQLAlchemy models and responsive Jinja2/Bootstrap views for authentication, dashboards, profiles and trek history.

FlaskSQLAlchemyJinja2Bootstrap 5SQLite

Built a production-ready multimodal conversational AI avatar with Tavus Conversational Video Interface (CVI) and LLM-based persona grounding, enabling real-time text, voice and live video interactions.
Developed a React + TypeScript frontend with live avatar streaming and a transcript UI.
Engineered a secure Node.js backend for conversation orchestration and REST API integration.
Implemented personalized knowledge grounding from structured Markdown for context-aware responses.
Designed scalable, low-latency real-time session management.

PythonReactTypeScriptNode.jsTavus CVILLM

Engineered an end-to-end deep learning pipeline that combines CNN feature extraction with Vision Transformers (ViT) for land-use classification.
Designed and trained custom CNN and ViT architectures from scratch, using data augmentation, regularization and dynamic early stopping to reduce overfitting by ~15%.
Tuned learning rates, batch sizes and optimizer schedules, speeding up training convergence by ~35%.
Reached 97.2% classification accuracy on the EuroSAT dataset, benchmarking convolutional backbones against transformer latencies with confusion matrices.

PythonPyTorchTensorFlow/KerasCNNViTData Augmentation

Built an end-to-end Azure pipeline in Azure Databricks and PySpark that automates ingestion, cleansing and transformation through Bronze, Silver and Gold Medallion layers, producing analytics-ready datasets.
Used Delta Live Tables (DLT) for orchestration and incremental processing, and Star Schema models with SCD Type 2 for historical tracking.
Enforced data governance with Unity Catalog and optimized the Gold layer for Power BI reporting.

AzureAzure DatabricksPySparkPythonDelta LakeDLTADLS Gen2

Designed a Power BI dashboard on official NPCI UPI transaction statistics to analyze and monitor India's digital payments ecosystem at scale.
Built a dimensional data model with reusable DAX measures, plus KPIs, trend analysis, MoM and YoY growth, rolling averages, cumulative metrics and transaction distribution insights.
Used Power Query for data preparation and delivered executive-ready self-service BI reports across multiple time periods.

Power BIDAXPower QuerySQLMicrosoft Excel

Want to talk?

Reach out about roles, projects or research.