Anirudh
Raghavendra

Processes 109 million real e-commerce events from an Eastern European electronics retailer across October–November 2019. Covers the full analyst workflow: raw data ingestion, a 5-step SQL preprocessing pipeline, funnel analysis, week-over-week trend tracking, revenue concentration, and brand efficiency — all visualised in a 3-dashboard Tableau Public report.

Segments retail customers using RFM analysis and clustering algorithms, then recommends tailored marketing strategies for each group — giving campaign teams a clear, data-backed plan of action.

End-to-end data science project analysing how the Pokemon competitive meta has shifted across all 9 generations (1996–2023), covering 1,194 Pokemon and 606 Smogon-tiered entries. Engineered a pipeline across 3 datasets, deriving 78 features per Pokemon including stat ratios, offensive/defensive splits, speed tiers, and 18-type one-hot encodings. Deployed as a 9-tab interactive Streamlit dashboard with a Pokemon-themed UI including sprite decorations, type badge icons, a Type Matchup Defender popup, SHAP waterfall charts, and filterable prediction tables.

A statistical experimentation framework built in pure SQL, processing 882,000+ rows across two experiments. Covers conversion rate analysis, Z-score and p-value significance testing, 95% confidence intervals with overlap detection, ad frequency segmentation across four exposure tiers, and day-of-week and hour-of-day breakdowns — all culminating in an executive summary report that flags winners and losers.

A system that observes your habits and automatically controls lights and thermostats to optimise comfort and reduce energy usage — without you having to program a single rule.

End-to-end ML pipeline predicting domestic flight delays using 1.86 million real BTS flight records across 2019, 2022, and 2023. Built a Python data pipeline that cleans, engineers features, and loads data into PostgreSQL — enriched with real historical weather data from the Open-Meteo API across 142 US airports, covering 62.5% of all flights. Integrates three data sources: BTS flight records, OurAirports coordinate database, and Open-Meteo weather API.

End-to-end HR analytics platform that predicts which employees are likely to leave, explains the reasons behind each prediction, and presents findings in an interactive dashboard built for non-technical HR teams. Three datasets merged (IBM HR Analytics, HRDataset v14, Employee Burnout) into a single enriched master dataset of 1,470 employees across 44 features — including engineered signals like burnout pressure index, promotion stagnation, and career velocity.

A natural language assistant for the 2024/25 Premier League season. Ask anything about any of the 20 clubs — results, standings, managers, stadiums, records — and get a direct answer with source citations. Built on a RAG pipeline with 21 embedded Wikipedia documents, query expansion, source prioritisation, and sliding window conversation memory for natural follow-up questions.

Upload any song and isolate its individual components — vocals, drums, bass, guitar — in seconds. Built for musicians, producers, and anyone who learns music by ear.