Sarfraz Hussain

I build applications and analyses that aim to be clear, reproducible, and useful for the decisions they support. The work here spans multi-agent AI systems, time-series forecasting, marketing and consumer analytics, business intelligence, and an early-stage product.

Featured projects

🤖 Generative AI

• Multi-Agent AI Marketing Strategy: a LangGraph pipeline of four AI agents with RAG and live web search, served through Streamlit
• Multi-Modal Claim Verification: Claude vision that decides whether claim photos support, contradict, or under-evidence a customer's claim, with a validated output schema and an evaluation harness

📈 Machine learning and forecasting

• Human Attention Forecasting: forecasts attention across platforms and content categories with ARIMA, Prophet, LSTM, and a Temporal Fusion Transformer, with a live Streamlit app
• Time Series Classification with ROCKET: multivariate activity recognition with a clean, leak-free cross-validation pipeline
• Salt Sales Forecasting: ARIMA versus machine learning for FMCG demand forecasting

📊 Analytics and research

• Ride-Hailing Marketing Analytics: regression, clustering, and scenario simulation to fix a ride-completion problem
• Power BI Dashboards: a star-schema sales dashboard and a Titanic survival explorer
• Coca-Cola PLS-SEM: two structural equation models of consumer purchase intention, built in SmartPLS

🚀 Product

• FreshZero AI: an expiry-intelligence prototype for grocery retailers that scores perishable stock by batch and recommends the smallest discount to recover sales before it expires

Toolbox

What I work on

• Generative AI: multi-agent systems, RAG, vision, and LLM application design
• Machine learning and forecasting: classification and time-series forecasting with ARIMA, Prophet, LSTM, and transformers, with reproducible evaluation
• Analytics and research: regression, clustering, segmentation, A/B testing, and PLS-SEM
• Data and product: dashboards, data modeling, and a React and Vite product prototype

Connect