Rahul Reddy Puchakayala · AI & Data Product Portfolio

Real AI.
Real Intelligence.
Real Results.

Three end-to-end AI & data systems powered by the Claude API. Every analysis, every segment profile, every experiment report is generated by a live LLM call — not hardcoded text. Paste your own customer data and watch real intelligence run on it.

● Claude API Live K-Means Clustering AI Segment Profiling NBA Recommendations Bayesian A/B Testing ETL + Identity Resolution AI Insight Reports
Project 01 · MarTech · AI/ML

AI Personalization Engine

Real K-Means++ clustering on customer behavioral data you control. Then Claude AI profiles each segment, reasons over a customer's vector, and writes a personalized next-best-action recommendation with real justification — live LLM call every time.

STEP 1 — K-Means++ Clustering on Real Customer Data Ready

Paste CSV rows (or use sample data). Format: name, spend, frequency, recency_days, digital_score, savings_score

K Segments 3
STEP 2 — AI Next-Best-Action Recommender (Live Claude Reasoning) Run clustering first

Select a customer. Claude sees their real feature vector, their segment profile, and all available message templates — then writes a live personalized recommendation with justification.

Select Customer
STEP 3 — A/B Test with Bayesian Analysis + AI Interpretation
Control CTR (%)4%
Treatment CTR (%)4.9%
Sample Size (each)1,500
Control (A)
—
Click-Through Rate
Treatment (B)
—
Click-Through Rate
Project 02 · Data Platform · Analytics

Customer 360 Data Platform

A real ETL pipeline that ingests from 4 source systems, resolves identities probabilistically, engineers features, computes Pearson correlations, and then calls Claude to write a live data quality report and targeting strategy — based on the actual computed output.

PIPELINE CONFIGURATION — 4 SOURCE SYSTEMS Idle
TRANSACTIONS
Records200
WEB EVENTS
Records180
CRM / EMAIL
Records160
MOBILE APP
Records150
Identity Match Rate87%
PIPELINE EXECUTION LOG
DATA QUALITY KPIs
Run pipeline to compute
UNIFIED CUSTOMER PROFILES — LIVE OUTPUT 0 profiles
IDLTV ScoreSpend 90dSessionsEmail OpensMobileChurn RiskPref. ChannelLifecycleSegmentDQ%
Run pipeline first
SEGMENT DISTRIBUTION
FEATURE CORRELATION MATRIX (Pearson r)
Run pipeline to compute
Project 03 · Product · Statistics

Experimentation Platform

Real sample size calculation using Cohen's h formula. Launch experiments, simulate days of traffic with real Z-test statistics, and when significance is reached Claude writes a live insight report — reasoning over the actual numbers, not a template.

EXPERIMENT DESIGNER — Real Sample Size Formula + Claude AI
Experiment Type
Baseline Rate (%)5%
Min Detectable Effect15%
Statistical Power80%
Significance α
Daily Traffic2,000
SAMPLE SIZE CALCULATOR (Cohen's h formula)
Required per variant—
Total sample needed—
Estimated runtime—
Cohen's h effect size—
Test typeTwo-sided Z-test
Primary KPICTR
Guardrail metricUnsubscribe Rate
LIVE EXPERIMENTS MONITOR
IDTypeDayCtrl RateTrt RateLiftN eachp-valueConfidenceStatusAction
No experiments yet
CONFIDENCE OVER TIME
CLAUDE AI — EXPERIMENT INSIGHT REPORT
Launch an experiment and simulate days.
When significance is reached, Claude writes a live report.