Real-time multilingual audio complaint processing system for the National Police Academy - transcribes, summarizes, extracts legal sections, and auto-fills FIR forms.

Secure, scalable system for archiving petabyte-scale satellite data at ISRO NRSC - with real-time monitoring, event streaming via Kafka, and automated notifications.

End-to-end OCR system and web platform for digitizing degraded historical Kaithi script documents - built for Bihar Government and Bhashini.

Pioneered GEO (Generative Engine Optimization) at Multilipi - the first capability of its kind in the product - enabling clients to be cited by ChatGPT, Gemini, and Perplexity.

Real-time website translation platform for Bhashini (MeitY) with a Maker-Checker quality assurance workflow for human translators.

Custom-built annotation tool for Kaithi script document images - polygon-level text region labeling with a Maker-Checker review workflow.

ML model to classify uninformative regions in satellite images, enabling intelligent data compression and prioritized transmission.

POC for generating Digital Elevation Models (3D terrain maps) from multi-spectral satellite imagery by fusing spatial data layers.

Land-use/land-cover classification and segmentation on multi-spectral satellite imagery using deep learning.

Time-series analysis and modeling of geospatial sensor data from satellite observations.

Pub/sub messaging system using Apache Kafka for high-throughput satellite data event streaming at ISRO NRSC.

POC for streaming massive satellite images directly into memory - bypassing disk I/O for faster processing and improved data security.

Core software architecture for coordinating a swarm of drones in real-time - near-zero latency comms and YOLO-based leader drone tracking.

Computer vision system to amplify and analyze micro-vibrations in video for equipment damage detection - 3rd Rank at Smart India Hackathon 2023 (National Level).

Volumetric deep learning system for segmenting tumor regions from 3D MRI/CT scan data using a 3D U-Net architecture.

Multi-class land-cover classification system using satellite imagery and deep learning - trained on multi-spectral data.