Turning messy data into decision-ready insight — and building the AI-assisted tool when one doesn’t exist.
Kolkata, India · Portfolio · LinkedIn · Credentials
| Now | Data & BI Analyst at Accenture |
| I work across | SQL & Python analytics · Power BI & Tableau · Statistical inference · Data quality · AI-assisted tool development |
| I build | Decision-ready dashboards, reproducible analytics pipelines, private-by-design desktop tools, and AI-enabled workflows |
| AI development | Claude Code · Antigravity · Codex · OpenCode · Cursor · GitHub Copilot |
| Credentials | Claude Code Certification · GCP Professional Data Engineer · Microsoft Power BI Data Analyst (PL-300) · Azure AZ-900 & SC-900 |
I pair analytical rigor with product thinking: the deliverable is not just a chart, but a clear decision and a workflow people can actually use.
I use AI to accelerate the work around analysis — framing an ambiguous question, profiling unfamiliar data, building repeatable SQL/Python workflows, and turning findings into clear documentation or usable internal tools.
The guardrails matter as much as the speed: metrics stay defined, assumptions stay visible, source data remains the authority, and conclusions are validated with data-quality checks, statistical reasoning, and domain context.
BUSINESS QUESTION → AI-ASSISTED EXPLORATION → REPRODUCIBLE ANALYSIS → HUMAN-REVIEWED DECISION
|
From raw transactions to executive action. A reproducible e-commerce analytics product with KPI reporting, RFM segmentation, cohort retention, ABC analysis, a Streamlit dashboard, and a static executive report.
|
Private NLP analysis for real-world text. A desktop application for clustering survey responses, support tickets, and customer feedback locally — without uploading sensitive data.
|
|
Know your queue before you report it. An offline desktop tool for auditing ITSM / ServiceNow ticket exports — catches silent data-quality issues (date ambiguity, state typos, missing fields, language mix) that Excel can't flag, and hands back a clean, reportable extract.
|
Analysis that explains its evidence. A reproducible study of 1,309 passengers using confidence intervals, odds ratios, effect sizes, an interactive dashboard, and a stakeholder-ready report.
|
|
The story behind the work. Case studies, analytical notes, and verified credentials — written for people who care how an answer was reached, not just what it says. Astro · MDX · TypeScript · GitHub Pages |
QUESTION → clean the signal → test the assumption → tell the story → ship a useful tool
I’m most effective where analytical detail needs to become a decision: defining trustworthy metrics, finding the cause behind a pattern, and making the result simple to use and explain.