KASAM — Multi-Sensor Situational Awareness

Defense / DRDO

AI-enabled platform integrating multi-sensor inputs, real-time dashboards, and autonomous decision support. Field-tested by the Indian Army.

Formal commendation from Armed Forces
Computer Vision Edge AI Real-time

SkyRoute — Helicopter Logistics Optimization

Defense / DRDO

AI-driven route & mission planner for rotary-wing logistics — optimizing paths under terrain, fuel, and operational constraints.

Open source on GitHub
Optimization Route Planning OR

RAG LLM Assistant for Operational Users

AI / LLM Research

Offline-capable RAG system for domain-specific Q&A and decision support — reliability, data isolation, and bandwidth-constrained deployment.

Offline-capable with full data isolation
RAG LLMs NLP

MARL + MANET Tactical Simulation

Defense Research

Multi-agent reinforcement learning testbed over mobile ad-hoc networks — adaptive coordination under contested, constrained network conditions.

Multi-agent coordination under network constraints
MARL Simulation MANET

Predictive Maintenance — WLR Swathi

Defense / DRDO

AI-driven predictive repair system forecasting failures and optimizing maintenance cycles for radar subsystems using historical and operational data.

Optimized maintenance for critical radar infrastructure
Predictive Analytics ML Radar

AI-Based Inventory & Logistics Optimization

Defense / Enterprise

Predictive analytics for inventory forecasting and logistics optimization in large-scale operational settings with multi-echelon demand patterns.

Large-scale operational deployment
Forecasting Logistics Analytics

Computer Vision Thermal Surveillance

Defense / DRDO

CV-based detection systems using thermal imagery for continuous monitoring in constrained field conditions — low-light, variable range, ruggedized.

Continuous monitoring in constrained environments
Computer Vision Thermal YOLO

Generative AI Research — Stable Diffusion

AI / Research

Text-to-sketch and sketch-to-image generation pipelines with a focus on controllability, fidelity, and deployment feasibility on modest hardware.

Controllable generation pipelines
Stable Diffusion Generative AI Text-to-Image

PPO in Matrix Games — Prisoner's Dilemma

M.Tech Research

Proximal Policy Optimization applied to classical matrix games to study emergent cooperation, policy convergence, and game-theoretic stability.

Open source on GitHub
PPO Game Theory Reinforcement Learning

CartPole — Deep Q-Network

RL Foundations

Classic DQN implementation on the CartPole control task — experience replay, target networks, and epsilon-greedy exploration built from scratch.

Open source on GitHub
DQN PyTorch OpenAI Gym

Amantranaa Caterers — Client Website

Client Engagement

Designed and delivered a production website for a catering business — full visual identity, responsive experience, and live on amantranaacaterers.com.

Live at amantranaacaterers.com
JavaScript Web Client Work

Timetable Generator

Utility Tool

Automated timetable generator that produces conflict-free schedules given faculty, room, and course constraints using constraint-satisfaction techniques.

Open source on GitHub
Python CSP Scheduling