Research that supports better products and stronger AI implementation.
The research track is practical: build knowledge that improves reliability, evaluation, modeling and deployment for real systems.
Active research directions
Each research area connects to real products, competitions, or implementation work.
Agentic AI
Tool-use agents, workflow automation and controllable AI systems
Research on agents that can use tools, generate reports, automate workflows, and remain testable, observable, and controllable enough for real business and research use.
AI Evaluation
Adaptive Testing of Learning Across Substrates
Research around measuring how AI systems learn, adapt, and fail under interactive conditions. This direction includes ATLAS Benchmark, replay-based diagnostics, and learning-focused evaluation.
Financial AI
Probability-based market decision systems
Research around feature engineering, thresholding, model packaging, risk-aware deployment, and financial NLP. This includes Golden Gauss AI for XAUUSD prediction and Short Activist Predictor for financial text classification.
Geo Spatial AI
Remote sensing, crop classification and satellite intelligence
Geospatial AI research inspired by crop classification, remote sensing, foundation models, and competition work. Combining satellite imagery analysis with machine learning pipelines.
Robust ML for Noisy & Imbalanced Data
Reliable learning under difficult data distributions
Exploration of machine learning methods for noisy, imbalanced, or complex classification settings, including hypersphere-based fuzzy SVM approaches and practical evaluation workflows.
Energy Flexibility AI
Demand response detection and building energy intelligence
Applied machine learning work on detecting and quantifying energy flexibility in buildings using a hybrid two-stage ensemble framework for classification and regression.
Publications & research work
Golden Gauss AI: A Predictive Machine Learning System for Probability-Based Trade Execution in Gold Markets
Published research on Zenodo. Covers feature engineering, gradient boosting, ONNX deployment, and MetaTrader integration for XAUUSD prediction.
GeoFM for Crop Classification
Geospatial foundation model approach for crop classification using NASA PRESTO, satellite time-series, and competition-grade ML pipelines.
Hybrid Two-Stage Ensemble Framework for Energy Flexibility in Buildings
Machine learning ensemble framework for predicting and optimizing energy flexibility in building systems.
ATLAS - Adaptive Testing of Learning Across Substrates
Benchmark suite evaluating how AI systems learn through interactive tasks. Covers associative learning, concept formation, probabilistic learning, and more.
Fine-Grained Animal Species ID Benchmark
A multimodal benchmark evaluating vision-language models on expert-level biological classification across animal species.
Startup Failures Dataset
Curated Kaggle dataset of startup failures across industries, funding stages, and regions. Built for analysis and machine-learning training.