All Projects.
A comprehensive directory of my engineering projects, AI model deployments, patents, and software explorations.
NeuroFusion-BCI
Lead AI Engineer
Proposed a novel Hybrid Feature Fusion methodology and engineered a sequential BiGRU + 1D CNN deep learning architecture to achieve robust spatial and temporal EEG pattern recognition for 4-class motor imagery tasks.
Key Accomplishments & Stack Integration
- Engineered sequential BiGRU + 1D CNN architectures, capturing complex temporal dynamics and spatial patterns of EEG signals.
- Developed a custom signal processing pipeline leveraging Hjorth parameters and bandpass filtering to maximize signal-to-noise ratio.
- Achieved 96.88% classification accuracy across 4-class motor imagery tasks, outperforming baseline models by 8.4%.
Saraav
Creator & Lead Developer
Launched a full-stack platform scaling to 500+ users. Implemented AI-driven content generation using Python scripts to automate study material creation. Managed the entire software development lifecycle (SDLC).
Key Accomplishments & Stack Integration
- Designed and deployed the entire database schema and cloud functions using Firebase Firestore and Auth.
- Developed an automated Python backend that leverages LLMs to generate high-quality practice questions and revision notes.
- Successfully scaled the web app to over 500 active registered users with real-time analytics and user progress tracking.
UAV Flood Mapping System
Patent: No. 202523064402 A
Designed an AI-enabled UAV system for real-time flood monitoring. Secured 2nd Prize at SGBAU University Level Avishkar Competition.
Key Accomplishments & Stack Integration
- Co-drafted and filed Indian Patent No. 202523064402 A for an integrated UAV-based aerial monitoring system.
- Trained custom semantic segmentation models (CNN-LSTM hybrids) to classify flooded regions in real-time from video feeds.
- Integrated GIS datasets and drone coordinates into a unified mapping dashboard for emergency response dispatchers.
Solar Panels Detection
AI/ML Engineer
Developed high-precision models to detect solar parks from ISRO's LISS4 sensor images for advanced environmental and asset mapping pipelines.
Key Accomplishments & Stack Integration
- Built semantic segmentation pipelines using U-Net architectures to detect large-scale solar arrays from high-resolution satellite imagery.
- Optimized dataset preprocessing for ISRO's LISS4 multispectral sensor data, handle class imbalance via focal loss functions.
- Exported model outputs into geospatial vector formats compatible with QGIS and ArcGIS pipelines.
Amazon Reviews Scraper & Analysis
Backend Engineer
Built a scalable web scraping engine to extract product reviews and analyze customer sentiment. Conducted data analysis utilizing LLMs (Claude/ChatGPT) to generate comprehensive product insights.
Key Accomplishments & Stack Integration
- Developed modular scraping agents using Selenium and Scrapy with Javascript rendering bypass via Splash JS.
- Built clean sentiment aggregation routines leveraging OpenAI and Anthropic API pipelines to summarize qualitative review trends.
- Containerized scraping engines in Docker, allowing orchestrations across multiple proxy networks to bypass IP bans.
