Case Studies
Eight deep-dives across quantitative research, supply-chain ML, e-commerce automation, banking data engineering, and applied AI research — each with hard numbers and production outcomes.
Building a Systematic Options Backtesting Framework
Investment Firm
Sharpe ratio of 4 in live index options strategies managing ₹100+ crore AUM, with full walk-forward validation and execution cost modeling.
Demand and Inventory Forecasting at Scale
Logistics & Supply Chain (SaaS)
7 production forecasting models deployed for a 15-member team processing 5TB+ supply-chain data across enterprise retail clients.
Cross-Border E-Commerce Intelligence for 1,000+ SKUs
E-Commerce Technology
50% manual effort reduction and 30% revenue estimation efficiency gain through ML-driven automation across 1,000+ SKUs on 4 marketplaces.
Automating Regulatory Data Pipelines at ANZ Bank
Tier-1 Australian Bank
50% reduction in development and testing time through end-to-end automation; production-grade pipelines processing regulatory data from 70+ source systems.
Temporal Graph Learning for Link Prediction
Academic Research
86% AUC on College Messages dataset — top performance using temporal attention over node2vec, TMF, CTDNE, and BANE baselines.
Crop Yield Forecasting for Hydroponics Systems
Agricultural Technology
Working ML pipeline integrating environmental, biological, and sales data for crop yield prediction with decision-support outputs for crop planning.
NLP-Driven Cancer Mutation Classification
Biomedical Research (Applied Project)
End-to-end ML pipeline for mutation classification with structured experimental discipline: 64/16/20 train/val/test split, class-balanced evaluation, and documented model comparison.
Semantic Duplicate Detection at Scale (Quora Question Pairs)
NLP Applied Project
End-to-end NLP pipeline combining hand-crafted semantic features with embedding-based similarity for duplicate detection, with documented ablation analysis.
Facebook Social Graph Link Prediction
Academic Project
AUC-ROC evaluation demonstrating that structural neighborhood features (Adamic-Adar, common neighbors) dramatically outperform node-level features for link prediction — a finding that directly informed the featurization strategy in subsequent NUS graph research.
Biomedical Second Opinion: ML Feasibility Study
Personal Research Project
Early-stage EDA revealed critical data quality barriers — annotation inconsistency across datasets and distribution shift between institutions — that would make clinical deployment significantly more complex than benchmark performance suggests.
Lets collaborate!
Whether you need a quantitative researcher, an machine learning systems builder, or a technical advisor — I'm available for select consulting engagements.
Get in Touch →