AKASH DOGRA

LEAD RESEARCHER / ENGINEER

Specializing in physics-informed neural networks, orbital mechanics simulations, and intelligent decentralized systems. Exploring the intersection of deep learning and mathematical physics.

H(q,p) = T(p) + V(q)

∇·E = ρ/ε₀

iℏ(∂Ψ/∂t) = ĤΨ

DOWNLOAD_SCHEMATICS.pdf INITIATE_CONTACT

MISSION LOGS

CMD-001

STABLE

Physics-Informed GNN Hamiltonian Neural ODE Control

Modeled large scale IoT networks as dynamic interference graphs using Neural ODEs. Incorporated Hamiltonian mechanics to enforce stable, energy-efficient dynamics.

CMD-002

ACTIVE

Dynamic Emergency Routing for Geo-Instable Regions

Real-time reinforcement learning based emergency routing system for Himalayan regions under dynamic hazard conditions.

CMD-003

ARCHIVED

Few-Shot Learning under Domain Shift

Contrastive few-shot learning framework with adversarial domain adaptation. Modified SimCLR to learn domain-invariant representations.

CMD-004

STABLE

DDoS Attack Detection in SDN Networks

Generated a large-scale labeled DDoS dataset using Mininet. Built an automated pipeline and achieved 98% detection accuracy using XGBoost.

CMD-005

ACTIVE

RL for Stock Trading Agent

Implemented DDPG, DDQN, and DQN for financial time-series decision-making. Modeled risk-aware performance using Sharpe ratio.