Projects built to solve real operational problems.

The problem

Organizations rush into AI adoption without a framework for deciding where it actually belongs. They invest in the wrong workflows, build tools nobody uses, and lose trust in the process. The gap isn't technical capability — it's structured decision-making at the front end.

What I built

Wolflow runs organizations through seven decision gates that surface whether a workflow is genuinely ready for AI — checking data availability, process stability, risk tolerance, and more — before anyone spends money or writes code.

Wolfpath takes it a step further: paste any workflow description and receive a prioritized adoption roadmap with effort estimates, impact scores, and sequenced next steps. It doesn't just identify opportunities — it tells you what to do first and why.

Stack

The problem

As organizations deploy LLMs internally, they face a structural problem: every prompt goes to the same model with no routing logic, no cost optimization, and no audit trail. When something goes wrong — or when compliance asks what the model was told — there's no answer.

What I built

Wolfprompt classifies every prompt across four dimensions — intent, sensitivity, complexity, and domain — then routes it to the appropriate model tier with a full decision log. The result is lower inference costs, better output quality for each use case, and an audit trail that compliance teams can actually use.

Stack

The problem

Enterprise AI platform features ship quarterly. Any static knowledge of what Copilot, Agentforce, or Now Assist can do is out of date within six months. But the question pattern is stable: given a platform and a problem, what can its AI layer do today, what does it cost to activate, and when does a custom agent outperform it? That's a retrieval problem, not a knowledge problem.

What I built

Wolfbridge runs live web search at generation time — nothing about platform capabilities is hardcoded. Users select up to four platforms from a grid of eight enterprise tools (Microsoft 365, Salesforce, ServiceNow, Google Workspace, and others), describe their problem, and receive a side-by-side briefing: what each platform's AI layer can do for this specific problem today, and two or three ranked custom-agent alternatives with build effort and cost estimates.

Every platform card includes a concrete limitation — what it cannot do for this problem. That's the detail that makes the output useful in a consulting context rather than a vendor brochure.

Stack

The problem

300+ vendor applications. A three-month manual review cycle. Data scattered across SharePoint and ServiceNow, requiring constant re-entry and reconciliation. The bottleneck wasn't people — it was structure. Information arrived in the wrong format, lived in the wrong place, and required human intervention at every handoff.

What I did

Mapped the entire workflow end-to-end, identified that the bottleneck was structural — not a capacity problem — and rebuilt data collection around structured SharePoint forms that automation could actually act on. Built Power Automate flows for risk categorization, approval routing, and ServiceNow ticket generation. Added executive dashboards for real-time visibility. Implemented validation rules at entry so errors never reached downstream systems.

Stack

The problem

Two project management systems — Jira for Agile teams, ServiceNow for Waterfall — created a data silo that made portfolio visibility impossible without manual effort. Executives needed status reports, but getting them meant three days of CSV exports, pivot tables, and reconciliation across platforms. By the time reports landed in inboxes, the data was already stale.

What I did

Sat in leadership meetings long enough to understand which questions executives actually needed answered — not which questions were easy to answer. Built API integrations pulling live data from both Jira and ServiceNow into a SharePoint data warehouse, with Power Automate handling scheduled refreshes and error recovery. PowerBI dashboards with department-level filtering, project health indicators, and one-click drill-down. The dashboard became the go-to resource for cross-departmental planning because it answered the right questions.

Stack

The problem

Healthcare providers lose billions annually to billing errors — gender-procedure mismatches, age-inappropriate treatments, anatomical inconsistencies — that slip through manual review. Traditional rule-based validation catches obvious violations but can't reason about clinical logic. It flags the error without explaining why it matters or what the downstream consequences are.

What I built

A validation platform combining deterministic rules with AI medical reasoning. Five parallel workers process claims simultaneously, with LRU caching for similar claim patterns. Each flagged claim gets a clinical explanation, business impact analysis, financial estimate, compliance assessment, and prioritized next steps — not just an error code.

Stack