Engineering work and projects

I think engineering credibility is shown by what you build, not by testimonials. Below are the products and developer tools I have built to solve real Power BI problems.

DAXQuest

Building an In-Browser DAX Practice Platform with Pattern-Based Answer Checking

Start practising DAX

Passive video tutorials give learners no feedback on the DAX they write. I built a self-contained browser platform with a Monaco editor and instant, pattern-based answer checking on a sample star-schema model.

01 // PROBLEM STATEMENT

Learning DAX requires developing an intuition for the evaluation context (row context vs. filter context vs. context transition). Passive video lectures provide zero feedback on subtle calculation mistakes (such as implicit context transitions in iterators or incorrect ALL/REMOVEFILTERS usage). Existing tools require firing up full Power BI Desktop with local tabular instances, creating massive friction.

02 // INVESTIGATION & ANALYSIS

I looked at what makes learning DAX hard. Learners need three things in one place: a responsive formula bar with syntax checking, immediate feedback on whether their answer is right, and a clear view of the sample model and filters each challenge assumes.

03 // ENGINEERING IMPLEMENTATION

I built DAXQuest as a Next.js application with a client-side DAX tokenizer and a challenge validator that compares normalized answers against accepted patterns and required functions. It does not execute DAX against data. I integrated Monaco Editor with Power BI DAX autocompletion, context-aware column suggestions, and interactive parameter hovers, and wrote 47 progressive engineering challenges spanning Aggregations, CALCULATE context transition, Iterators, Time Intelligence, and Capstone Sales Analytics.

04 // MEASURED OUTCOME

A zero-install practice platform that checks DAX answers instantly in the browser. It includes 47 challenges, 20 lessons, an automated diagnostic assessment, and a comprehensive DAX reference guide.

Stack:Monaco EditorAnswer CheckingStar SchemaDAX PracticeTypeScript

BIKit

Standardizing Developer Workflows with PBIP and TMDL Tooling

Explore BIKit Tooling

Monolithic binary .pbix files prevent multiple BI developers from collaborating on the same model, reviewing changes in pull requests, or automating CI/CD pipelines. I built CLI utilities to serialize, lint, and validate Tabular Model Definition Language (TMDL).

01 // PROBLEM STATEMENT

For enterprise teams with 5+ BI engineers, the binary .pbix file is a bottleneck. Merging two branches is impossible; files overwrite each other; measures break silently in production; and code reviews happen post-deployment. Teams waste hours manually synchronizing measure definitions.

02 // INVESTIGATION & ANALYSIS

With Microsoft's release of the PBIP format and Tabular Model Definition Language (TMDL), semantic models became plain-text files. However, tooling to enforce formatting standards, measure documentation, and automated deployment checks was missing.

03 // ENGINEERING IMPLEMENTATION

I designed BIKit as an engineering CLI and GitHub Action runner that inspects TMDL files. It validates DAX formatting standards, flags undocumented measures, detects orphan columns, and verifies foreign-key integrity prior to merging pull requests into the production branch.

04 // MEASURED OUTCOME

Teams utilizing PBIP + BIKit operate with true Git pull request workflows. Merge conflicts are resolved as plain text, measure regressions are caught in CI, and deployments to Fabric workspaces occur automatically.

Stack:TMDLPBIPGit WorkflowsCI/CDAutomated Linting

powerbi.dev

Automated Semantic Model & VertiPaq Cardinality Diagnostics

Explore powerbi.dev

Enterprise models accumulate hidden high-cardinality columns, snowflake joins, and bidirectional filter loops that inflate memory footprints from hundreds of megabytes to multi-gigabytes. I created automated diagnostics to profile VertiPaq dictionary sizes and generate actionable remediation roadmaps.

01 // PROBLEM STATEMENT

Organizations frequently run out of Power BI Premium capacity memory without knowing why. Developers often leave transaction IDs, system GUIDs, and high-precision timestamps in the model, unknowingly consuming critical capacity RAM while visuals experience calculation lag.

02 // INVESTIGATION & ANALYSIS

Profiling enterprise models showed that the majority of memory bloat typically comes from a small set of high-cardinality columns that no user visual ever queries. Furthermore, bidirectional relationships between dimensions created costly storage engine callback loops.

03 // ENGINEERING IMPLEMENTATION

I built diagnostic heuristics that inspect semantic model metadata (BIM, TMDL, and VertiPaq Analyzer metrics). The tool automatically scores models across 10 engineering criteria: column cardinality, relationship ambiguity, DAX iterator usage, auto date/time table overhead, and Fabric Direct Lake readiness.

04 // MEASURED OUTCOME

Automated diagnostic reports that pinpoint high-cardinality columns, identify uncompressed dictionaries, and flag non-folding M steps before deployment to production capacities.

Stack:VertiPaqCardinalityDiagnostic EngineStatic AnalysisMemory Profiling