Skip to content
Back to portfolio

ML Lab

Machine Learning Models

Eight end-to-end models, each applied to a different domain from my IIT Madras coursework and productivity workflows — classical classification, deep-learning sequence models, document intelligence, and real-time behavioral syncing.

Python · scikit-learn · TensorFlowGoogle Gemini · Supabase · FirebaseHuggingFace SpacesMongoDB AtlasGradio · Streamlit
01Live

Construction Cost-Overrun Predictor

Construction Economics

Predicts budget-overrun severity from project economic and contractual parameters. Every prediction is logged to MongoDB.

Tier 1 — Foundations

View →
02Live

Credit Risk Classifier

DSAI in Finance

Classifies loan applicants by default risk using ensemble methods. Includes SMOTE for class imbalance.

Tier 2 — Ensembles

View →
03Live

Construction Quality Defect Flagger

Construction Quality

Unsupervised anomaly detection on quality inspection data — no labelled defects required to train.

Tier 3 — Anomaly Detection

View →
04Live

Hazard Image Classifier

Construction Safety

Detects safety-hazard categories in construction site images using a CNN trained on labelled examples.

Tier 4 — Deep Learning

View →
05Live

Power Load Forecaster

Smart Power Grid

Forecasts electricity demand using an LSTM that learns the temporal patterns in historical load data.

Tier 5 — Sequence Models

View →
06Live

SCADA Anomaly Detector

EMS SCADA

Detects deviations from normal SCADA operating behaviour using reconstruction error from a trained autoencoder.

Tier 6 — Unsupervised Deep

View →
07Live

Database Indexer

Municipal Infrastructure

Automates the digitization of handwritten complaint and maintenance forms for the Karnataka Urban Water Supply Modernization Project.

Tier 7 — Document Intelligence

View →
08Live

Habit Forge

Skill & Behavior Tracking

Tracks habits daily, syncs real-time with Firestore, and copies structured logs formatted for direct pasting into Google Sheets.

Tier 8 — Behavioral Intelligence

View →