Fluid Intelligence
in Motion
AI & Reinforcement Learning Engineer. Designing emergent multi-agent self-play systems, deep computer vision, and cryptographic steganography.
Harmonizing Complexity with Luminous Clarity
Architectural depth spanning probabilistic reinforcement learning agents, perceptual deep neural networks, and fault-tolerant cloud pipelines.
Autonomous Agents (PPO/MADDPG)
Self-play multi-agent dynamics with continuous Pymunk physics, GPU parallelization, and custom reward engineering. Optimized for adversarial and cooperative policy stability.
Computer Vision & Security
Aesthetic steganographic QR codes utilizing Stable Diffusion and ControlNet with distance-authenticated cryptographic signatures and generative artifact mitigation.
Cloud & Predictive AI
Enterprise time-series forecasting (ARIMA) deployed to auto-scaling IBM Cloud environments with automated inference drift monitoring and container orchestration.
Fellowships & Engineering Leadership
Ministry of Electronics and Information Technology (MeitY)
AI and Security Research Intern
Pioneered generative error-corrected QR steganography using Stable Diffusion and ControlNet for defensive covert communications. Formulated mathematically sound loss bounds balancing perceptual quality and barcode decodability.
Edunet Foundation in collaboration with IBM
AI and Cloud Intern
Engineered predictive ARIMA time-series models and deployed production-grade REST endpoints on IBM Cloud infrastructure with containerized pipelines for high-throughput automated forecasting.
ACM MUJ Student Chapter
Head of Projects and Research Team
Directed 15+ student developers selected from 40+ applicants, maintaining a 90% project completion benchmark through iterative agile code reviews, architecture critiques, and sprint cadences.
Selected Implementations
High-performance distributed learning systems, multi-agent coordinate networks, and automated synthesis engines.
MARL Tag: 3v3 Pursuit-Evasion
PyTorch self-play reinforcement learning system utilizing PPO with Generalized Advantage Estimation (GAE). Parallelized over 8 CUDA environments (30M steps) with 102-ray perception and 95% victory rate against algorithmic baselines.
Multi-Agent Fleet Transit Optimizer
Autonomous taxi simulation combining Multi-Agent Deep Deterministic Policy Gradients (MADDPG) with dynamic A* pathfinding. Slashed passenger wait times by 20% across 100+ simultaneously simulated agents in congested grids.
Student Performance Diagnostic
Random Forest classifier achieving 96% evaluation accuracy; optimized via rigorous hyperparameter tuning (GridSearchCV), feature importance factorization, and multi-variable normalization for proactive educational intervention.
Automated Video Generation Engine
End-to-end Python engine converting structured metadata and real-time feeds into programmatic video assets using MoviePy and REST endpoints, reducing manual rendering latency by 90%.
Computational & Engineering Matrix
code Languages & Core Systems
psychology AI, Frameworks & Environments
Connect with Aryan Gupta
Available for 2026 roles in AI, Machine Learning, and Systems Engineering. Whether discussing emergent multi-agent reinforcement learning topologies or production deployment pipelines, reach out directly.