A collection of applied AI, data, research and software work.
Organised around the problem, technical decisions, evidence and limitations of each project—rather than around tools or buzzwords.
LenseESG — Explainable Multimodal ESG Risk Intelligence
An end-to-end research pipeline combining structured ESG indicators, company disclosures and news-derived signals to create more interpretable company-risk assessments.
Deep Learning for Cyclone Track Forecasting
A Springer-published comparative study of LSTM, GRU and ConvLSTM models for forecasting cyclone trajectories from historical Bay of Bengal data.
Streaming Anomaly Detection for Cyber-Physical Systems
A complete monitoring workflow that connected simulated sensor data, anomaly-detection models, time-series storage and live dashboard reporting.
Data Reconciliation and Operations Monitoring Dashboard
A sanitised operational analytics case study that reconciled records across two source systems, identified missing encounters and turned exceptions into an actionable dashboard.
Transformer-Based Stack Overflow Question Classification
A BERT-based workflow for categorising unstructured technical questions to support routing, search and knowledge management.
Spam SMS Classification and Cloud Deployment
A text-classification workflow that moved beyond notebook modelling toward execution and comparison across Ubuntu and OpenStack-style environments.
Facial Expression Recognition
A CNN-based image-classification workflow covering seven facial-expression categories.
Fraud Detection and Model Benchmarking
A comparative classification workflow for identifying suspicious or anomalous transactions using preprocessing, model benchmarking and error analysis.
A/B Testing and Website Variant Delivery
Implementing and quality-checking frontend website variants for controlled experimentation as part of an engineering internship.
Recommendation-System Experiment
Explored user–item modelling and recommendation fundamentals to understand how behavioural data can support ranking and personalisation.
AI-Enabled Business Software Concept
A low-code AI-assisted software concept for coaches and small businesses—mapping client management, reminders, scheduling, FAQs and social-media support into practical workflows.