Tempe, Arizona · open to relocation
I build the pipeline, the model, and the dashboard that sits on top of them — and I bring the baseline that proves the model was worth building.
Data Analyst
Community Dreams Foundation
Four years turning messy source systems into reporting people actually run their week on — SAP extracts, healthcare referral data, student cohorts, warehouse telemetry.
Most of my work starts where the data is still wrong. I write the SQL and the Python that cleans and validates it, model it into something a BI tool can query without a fight, and then build the Power BI or Tableau layer stakeholders open every morning. At Vaibhav Laxmi that meant an ETL pipeline over a million SAP records with a data-quality framework in front of it; at Community Dreams Foundation it means Microsoft Fabric and OneLake feeding referral and rehabilitation dashboards for a care-coordination platform.
I care about the part most dashboards skip: whether the number is defensible. Every model I ship gets a baseline above it in the table — if a forecast can't beat "same weekday last week", it hasn't earned its place in a planning meeting. That habit is the through-line between the analyst work and the machine-learning work, and it's what the two projects below are built to demonstrate.
These titles overlap more than job postings admit. Here is the concrete evidence behind each one.
Writing and optimizing the queries, cleaning what comes back, and answering the ad hoc question before the dashboard exists. Built validation and anomaly-detection frameworks that raised data reliability to 93% and cut 8+ hours of weekly manual QA.
Sitting with stakeholders to turn "we need better visibility" into a defined metric, an owner and a refresh cadence. Gathered requirements across operations, finance and faculty teams and translated them into reporting specs and process changes.
Dimensional models, DAX measures that survive contact with a filter, and dashboards that load fast enough to be used. Power BI and Tableau in production across healthcare, manufacturing and education contexts.
Forecasting, clustering, recommendation and NLP — always benchmarked against the naive rule the business already uses, with the split designed so the model can't see the future it's predicting.
Both are complete repositories — clone, run one command, and every figure in the README regenerates. The dashboards below are the live output.
Python · SQL · Star Schema · Scikit-learn · Tableau-Ready Extracts
A four-node fulfilment network end to end: a relational WMS extract and a document store of device telemetry and free-text shift notes, consolidated into one star schema behind an 11-rule data-quality gate. On top sit a workload forecast, an NLP layer over supervisor notes, and a single-file dashboard.
The result worth arguing about: unsupervised clustering of the notes was weak (ARI 0.15) while a supervised classifier hit 98.4%. The repo says so plainly and recommends bootstrapping a taxonomy with clustering, then shipping the classifier — reporting the honest number was more useful than burying it.
Python · NumPy/SciPy · Scikit-learn · Collaborative Filtering · K-Means
Nine recommenders benchmarked on one split under one protocol — popularity and bias baselines, item-item and user-user neighbourhood CF, truncated SVD, regularized matrix factorization, a content model and a tuned hybrid — all implemented on NumPy and SciPy rather than called from a library. Plus five behavioral reader segments and a model predicting next-quarter reading volume.
Two findings the README leads with rather than hides: truncated SVD — the version most people reach for first — loses to a neighbourhood model, while properly regularized matrix factorization wins outright; and the tuned hybrid improves NDCG by 0.0007, which is a component that did not earn its place.
Three production dashboards from healthcare, manufacturing finance and education. These run on internal data, so what follows is what each one measures and why — not a screenshot.
Community Dreams Foundation · Power BI, Microsoft Fabric, OneLake, MySQL, Azure Data Studio, Python
The reporting layer for a care-coordination platform. Referrals arrive from multiple partner systems and have to be tracked from intake through placement and rehabilitation outcome — a funnel where a stall at any stage is a patient waiting. The dashboard replaced a set of manually assembled extracts with a Fabric-fed model that refreshes on schedule.
Vaibhav Laxmi Spinning Mills · Power BI, SAP, Python, SQL, ETL
Built on an ETL pipeline over 1M+ SAP records, with the data-quality framework that raised reliability to 93% sitting between the extract and the model. The dashboard served two audiences at once — an executive view for the monthly review, and a working view the inventory and finance teams used daily — which cut reporting turnaround sharply and removed 8+ hours of manual QA a week.
ASU Learning Enterprise · SQL, Tableau, Excel, HubSpot
Supporting 300+ students across three technical courses meant the grading queue and the support queue both needed to be visible. This tracker pulled assignment outcomes and HubSpot query volume into one place so faculty could see where a cohort was struggling while there was still term left to do something about it.
These dashboards contain proprietary and, in the healthcare case, protected data. They are described here rather than shown. I'm happy to walk through the structure, the modeling decisions and the DAX in a conversation.
Master of Science, Information Technology — GPA 3.87
Coursework that maps directly onto the work: Analyzing Big Data · Advanced Big Data Analytics & AI · Advanced Database Management Systems · Data Visualization & Reporting for IT · Natural Language Processing for IT · Data in the Cloud · Advanced Information Systems Security
Jan 2024 – Dec 2025 · Tempe, AZ
Bachelor of Technology, Computer Science & Engineering
May 2018 – Jul 2022 · Ahmedabad, India
Everything a hiring process asks for, one click away. Each button writes the email for you — add your name, hit send, and it comes back the same day.
One page: experience, projects and the full technical stack. The same content as this page, formatted for an applicant tracking system.
Sent the same dayFrom Joey Choi, CIS Online Course Manager at ASU Learning Enterprise, who supervised my grading work across the cloud and big-data courses.
Sent the same dayMS in Information Technology, Arizona State University — 30 credits, cumulative GPA 3.87.
Sent the same dayManagers and faculty who supervised my work at Community Dreams Foundation, Vaibhav Laxmi Spinning Mills and ASU Learning Enterprise. Names and contact details go out on request, and I let each referee know to expect you.
Sent the same dayRequests go to jshah96@asu.edu with the subject line already filled in, so I know exactly what you need. Documents are sent directly rather than posted publicly — transcripts and references carry personal details that belong in an inbox, not on a web page.
Open to Data Analyst, Business Analyst, BI Analyst and Data Scientist roles — in Tempe, remote, or wherever the team is. Fastest way to reach me is email.