One engagement, eight workstreams.

Problem

A 300+ location operator scaling toward 1000+ — competing with national chains like Caliber and Gerber — was running on an on-prem data warehouse it had outgrown. It could not respond to data needs fast enough to feed an acquisition-driven growth plan.

Approach

Architected and led the migration to a cloud-native stack — Azure, Snowflake, dbt, and Qlik — with webhook and event-based ingestion, near-live incremental modeling, and CI/CD plus observability built in. Grew the engagement from 2.5 to 10+ FTE over three years.

Result

Under 15 minutes end-to-end for nearly all flows, near-live reporting across every key metric, and a foundation that gets more cost-efficient as volume grows.

Problem

Insurance carriers send electronic-payment details that staff allocated to each repair by hand — slow, costly, and impossible to scale, with a manual data-entry step into the estimating software on top.

Approach

Automated the match programmatically off the operations data, used external API calls to automate the data entry, and paired it with a Qlik observability app so the team could manage the exceptions instead of the whole queue.

Result

About 97% of electronic-payment allocations now run automatically, cutting significant labor and software-licensing cost, and the process scales cleanly as the business grows.

Problem

The accounting integration ran on a packaged connector (Celigo) that was cumbersome to change and left finance with little control over how transactions flowed into NetSuite.

Approach

Built a custom Snowflake-to-NetSuite pipeline — SOAP requests via Snowflake External Access, Azure Data Factory for ingest and modeling — syncing transactional sales data in near real-time.

Result

Finance now owns its AP, AR, and month-end workflows, the sync runs near real-time, and the pipeline adds anomaly detection and observability the old connector never had.

Problem

Moving HR off ADP onto UKG, the business needed near-live HR data to drive automations — but UKG offered no push architecture to get it.

Approach

Built a custom event-driven pipeline using the Azure Durable Functions "monitor" pattern against UKG's table-and-field-level audit log, replicating it in Snowflake to mint events for the specific table+field pairs that mattered.

Result

Near-live HR data across systems, with employee events automatically opening ITSM tickets for onboarding and offboarding.

Problem

Poorly written repair estimates damage the operator's relationships with insurance carriers — and its revenue. Auditing every file through internal "Central Review" was inefficient and hard to customize.

Approach

Built an Azure Functions API that receives each estimate as a webhook and decides, in under 15 seconds, what to do with the file — applying custom scoring rules pulled live from Snowflake. A Streamlit UI lets the business own and tune the scoring logic.

Result

Estimates are scored in real time, only the files that actually need review are routed to Central Review, and the business controls the rules without waiting on a developer.

Problem

Operators across many locations have to prioritize which conversations to have and where to travel — but they aren't, and shouldn't have to be, data analysts staring at dashboards.

Approach

Built an ML anomaly-and-trend detection pipeline over a governed KPI tree — Sales, Quality, Speed, CSAT — with pre-calculated change detection across KPIs at multiple levels, then a multi-agent layer on top. Ask about a top-level metric like NPS and it automatically surfaces the drivers underneath it, like communication response time and promise-date accuracy.

Result

Governed, actionable insight for a specific location instead of another dashboard or spreadsheet — and it flags what is actually shifting, week to week.

Problem

Lower-tech middle management needed a simplified, actionable view; the full Qlik reporting repository was more than they wanted to navigate.

Approach

Built a role-based landing page that combines "must-see" objects with individualized KPIs. KPI data is normalized and ranked — top 3 and bottom 3 with geographic ranks — and the page doubles as a central jumping-off point for support needs like the help desk and HR.

Result

A simplified, actionable hub field staff actually use, with Qlik still underneath as the deep reporting repository.

Problem

Estimators add, remove, and change lines inconsistently, which drives carrier change-requests and rework downstream.

Approach

Built an estimate-recommendation model with SDG's global Business Science vertical, analyzing historic estimates by damage type, vehicle, and carrier and using association rules and anomaly detection to recommend which lines to add, remove, or change.

Result

Recommendations that cut change requests by catching and correcting estimates before they go out.