-- data analyst · troy, michigan
I turn messy operational data into reporting that leaders trust, with SQL, Python, Power BI and Tableau. 5+ years across automotive, consulting and construction.
Five roles, three industries, one habit: check every number back to source.
Jan – Jul 2026Auburn Hills, MI
Stellantis (Contract via Methodica Technologies)
Jan – Apr 2025Troy, MI
Methodica Technologies
Oct 2024 – Feb 2025Troy, MI
Walsh College
Jan – Aug 2024Pune, India
Bright Consultancy Services
Jan 2020 – Dec 2023Belagavi, India
Shree Samarth Constructions
The question, the method, and what the numbers said.
NSW remote-work surveys (3,019 responses) plus a 1,000-worker dataset. Remote workers scored higher on productivity and well-being (p < 0.0001), but remote share alone explained little (R² ≈ 0.02), so the call was a structured hybrid model.
X̄–R charts showed crust thickness in control, so the defects were in order-taking: two error types made up 78% of errors at 49,444 DPMO. Fix: address validation and time-stamping at the POS.
Designed the study (survey, n = 384 via Cochran's formula, paired t-test and regression) and tested it on simulated store data: +20.7% mean monthly sales, p < 0.01.
Compared eight scikit-learn classifiers on 3,276 samples. Best model (SVM) reached 67% against a 61% baseline; features barely correlate with potability (|r| ≤ 0.04), which caps any model. GitHub
Strategy, websites and proposals for a Bloomfield Hills sign company and its owner's music label.
Five-page site (HTML, CSS, vanilla JS) with services, story, a 36-photo gallery and contact, built from the client's own copy and project photos.
Music catalogue with 30-second previews, "listen on" links to the major streaming platforms and a licensing section for film and TV supervisors.
Strategy & Operations Lead · 2026
Work, internship and coursework.
Wrote the dashboard requirements and built Grafana views on InfluxDB test-run data: execution summary by test bench, errors by exception type, and system utilization.
ESP8266 and DHT11 readings streamed to Google Sheets for remote monitoring.
Hybrid Random Forest, XGBoost and autoencoder framework on 5,000 synthetic transactions, with SMOTE for imbalance and SHAP and LIME for explainability.
Pivoted 5,000 loan records and showed loan attributes do not predict jobs created.
ERD-to-SQL relational design with associative tables and indexed analytical queries.
Cost vs efficacy from cited veterinary sources: ranitidine had the lowest cost per efficacy point; omeprazole reached 85% at 7.8x the cost.
Linear, Ridge and Lasso regression on the Kaggle House Prices dataset.
Choosing the right chart for each audience: trends, seasonality and pollutant drivers.
Word frequency and dictionary sentiment, first built in Excel, now running on the full novel.
EC2 web servers behind an Elastic Load Balancer with health checks.
Python to Power BI on public data: EV charging reliability, vehicle recalls, freight and energy forecasting, data-quality audit.
6Cohort retention, funnel analysis on BigQuery, A/B test readout, KPI one-pager, SQL library, ETL pipeline.
6Segmentation, churn prediction, statistics in R, Tableau, advanced Excel and Grafana monitoring.
Skills, education and certifications.
-- let's talk
On-site, hybrid or remote. Scan to save my contact or reopen this portfolio.