Forward Deployed Engineer at Microsoft · Reno, NV
Greg Chedwick
Domain agnostic: I build the data models, then the dashboards and automation people actually use.
Senior engineering and analytics professional with 20+ years in data analytics, 6+ of them leading initiatives, across software licensing, mortgage and consumer lending, advertising, supply chain, and freight. I engineer scalable solutions — data modeling, transformation, warehousing, and ETL pipelines in SQL and Python — then build the dashboards and AI-driven automation that sit on top of them, embedded with the teams that depend on them and treating data quality and privacy as requirements rather than afterthoughts. I do my best work in ambiguous 0-to-1 territory with minimal oversight.
By the numbers
Impact, quantified
Every figure below traces to a specific project in the career history — no round numbers without a story behind them.
4,300+
Hours automated per year
Anniversary and mid-term ordering workflows, enabling 3x business scaling in two years
$106M
On-time renewal lift
Driven by compliance analytics and automated deep-dive reporting
$3.2B
Agreement portfolio in view
Power BI dashboards giving stakeholders live visibility into portfolio health
$31M
Cost reduction
Data-driven business cases built with DMAIC and Agile methodologies
$5.7B
Revenue enabled
Process and system improvements substantiated through analytics
$60B+
Portfolio remediated
Loan modification campaigns supporting $25B+ in government programs
Career
23 years, three employers
Each bar spans one employer. Select a role to read what it involved — or read them all below, no clicking required.
Microsoft
Forward Deployed EngineerAug 2026 – PresentFrontier AI, Business Operations
- Embedded with business teams to design and ship a Dataverse-based intake and workflow platform — automated routing, approvals, and audit history — replacing a fragmented manual process and cutting over in-quarter.
- Root-caused a deployment defect that silently overwrote production configuration on every release, corrupting downstream data with no error surfaced; scripted the remediation across 11 broken connections and restored integrity without a rollback.
- Built AI-assisted automation to synchronize work-tracking state and delivery alerts across systems, replacing recurring manual triage with an automated pipeline.
Senior Data AnalystAug 2022 – Jul 2026Fusion Development, Operations Service Center
- Engineered scalable licensing data models, automated workflows, and developed Power Apps that automated anniversary and mid-term ordering — saving 4,300+ hours annually and enabling 3x business scaling over two years.
- Developed compliance analytics metrics and Power BI dashboards giving visibility into a $3.2B+ agreement portfolio, driving a $106M increase in on-time renewals.
- Designed agreement complexity models to identify bottlenecks, enabling targeted root-cause investigations and scalable automation for ad-hoc requests.
- Built custom AI automation with Copilot Studio to enable stakeholder self-service, integrate disparate data sources, and accelerate insight delivery.
Business Analytics SpecialistAug 2018 – Jul 2022Business Process & Analytics, Commercial Ops
- Led global analytics backlog prioritization using Cost of Delay / Weighted Shortest Job First within a SAFe framework, ensuring timely deployment of high-impact BI across software licensing, advertising, and supply chain.
- Partnered with cross-functional stakeholders to define requirements and ship scalable dashboards that informed decisions and supported new program launches.
- Developed automated reporting pipelines and self-service tools across digital attach, hardware compliance, search, and advertising.
Business Operations AnalystJun 2015 – Jul 2018Process Management, Commercial Ops
- Crafted data-driven business cases using DMAIC and Agile methodologies, substantiating improvements that cut costs by $31M, reduced AR exposure by $400M, and enabled $5.7B in revenue.
- Engineered BI solutions and monitoring dashboards to track outcomes of process and system improvements.
Bank of America
Vice President, Consumer Products Strategic ManagerAug 2010 – May 2015Loan Loss Mitigation and Portfolio Analytics
- Led the analytics team behind loan modification campaigns, remediating a $60B+ portfolio and supporting $25B+ in government programs through advanced BI and reporting.
- Managed BI strategy, reporting, and operational process for Enterprise Complaint Resolution, covering mortgage, credit card, and deposit escalations from state attorneys general, elected officials, the CFPB, the OCC, and other regulators.
- Presented program performance, risk trends, and operational insight to executive leadership, advising on initiatives that improved complaint resolution efficiency and regulatory compliance.
- Directed production of the National Mortgage Settlement Servicing Standard Scorecard for executive review and submission to the Office of Mortgage Settlement Oversight.
- Oversaw reporting and process development for the U.S. DOJ National Mortgage Settlement Program, enabling modification of $25B+ in loan balances.
- Managed a team of 4 senior analysts and coordinated 8 operational managers across loan modification and complaint resolution programs.
- Built and automated an enterprise-grade BI and data warehouse supporting marketing, production, modeling, forecasting, and operational decision-making across proprietary and government programs.
- Analyzed portfolio trends across risk, losses, delinquency roll rates, payment shock, rate resets, and FICO vintage, alongside campaign effectiveness, and delivered the findings to executive leadership.
- Led cross-functional work with finance, technology, marketing, legal, compliance, and executive management to optimize program performance and regulatory adherence.
- Directed UAT, training, and deployment of loan modification and refinance tools across multiple business lines.
Charles Schwab Bank
Finance Manager, Bank FinanceAug 2003 – Aug 2010Bank Finance
- Built and administered enterprise loan, sales, and marketing databases supporting Finance, Credit, Compliance, and senior leadership.
- Led analytics infrastructure modernization, serving as subject matter expert for the MS Access to SQL Server migration.
- Enhanced and managed loan pricing models, and executed pricing exception decisions using ROE, spreads, cost of funds, LTV, FICO, and credit history.
- Produced executive-level reporting for the Board of Directors, ALCO, and the Credit Risk Management Committee.
- Designed delinquency roll-rate, industry comparison, and portfolio vintage analytics to assess performance and risk trends.
- Developed industry rate databases and analyses guiding Pricing Committee lending decisions.
- Built marketing effectiveness analytics enabling strategic campaign decisions for the Chief Marketing Officer.
- Created streamlined pricing exception processes, improving the speed and consistency of lending decisions.
- Managed junior analysts to ensure accurate, timely reporting across Finance, Credit, Accounting, Marketing, Legal, and Compliance.
- Coordinated regulatory reporting for the Office of Thrift Supervision and supported OCC annual review analytics.
Toolkit
Skills, by years in the seat
Hover any tool in the ticker for what I've done with it. Below, bar length is hands-on years on one shared scale — two entries are marked as recent rather than deep, because being straight about that matters more than a longer bar.
Data & BI
- SQL15+ yrs
Complex querying, data modeling, large-scale analysis
- KQL5+ yrs
Kusto queries over log and telemetry data
- Power BI & DAX10+ yrs
Interactive dashboards, reporting, visualization
- SSMS & VS Code10+ yrs
Data modeling, analysis, AI-assisted coding with Claude Code and Copilot CLI
Data Engineering
- Microsoft Fabric & SSIS10+ yrs
ETL pipelines, SQL databases, lakehouses
- Azure DevOps10+ yrs
Plan, build, test, and deploy solutions
- Azure Platform Services5+ yrs
Function Apps, Logic Apps, Data Factory, serverless automation
- PythonGrowing
Data manipulation, scripting, automation
Automation & AI
- Power Platform7+ yrs
Power Apps, Power Automate, Dataverse, solution ALM and deployment pipelines
- Microsoft Copilot StudioRecent
AI agents, low-code intelligent automation
Education
Master of Business Administration
University of Nevada, Reno · 2013
GPA 4.0 · Data Resource Management, Information & Communication Technology, Strategic Management
B.S. Business Administration
California State University, East Bay · 1998
Minor in Computer Science
Nights & weekends
Projects
Things I build outside work to stay sharp on the tools. Each one ships end to end — data in, decision out — rather than stopping at a notebook. Pick a card for the full analysis.
Problem framing, data forensics, feature engineering, modelling, validation
Will a trucking carrier still be operating a year from now? Brokers, insurers and lessors all carry exposure to carriers that may not exist at renewal, and every public tool is backward-looking. This model scores 1.9M carriers from public FMCSA records. The figures below are the real evaluation output, not illustrations.
- 0.889AUCagainst 0.748 for a two-column baseline
- 6.7xLiftover a 2.2% base rate
- 1.9MCarriers11.8M carrier-months, 26 features
- 0.0008Calibration errorpredicted risk tracks reality to ~0.08pp
What a review queue actually buys
Share of all failures caught when carriers are reviewed highest-risk first.
The thin rule on each bar marks where random selection would land. Reviewing one carrier in ten surfaces 58% of everything that fails in the following year.
Holds across fleet sizes
AUC by power units. 0.5 is a coin flip.
Small carriers are most of the population, so they are segmented at evaluation rather than filtered out of training — dropping them would move the base rate, not improve the model.
The hard part was the label, not the model
The obvious target — “did FMCSA revoke this carrier’s authority?” — is wrong, and the data says so plainly. Of the revocation proceedings on record, 2,208,586 were discontinued against 1,600,771 executed. Most begin with a lapsed insurance filing; the carrier files a new policy and keeps trading. Training on “a revocation happened” means training mostly on carriers having a bad month.
Exit is therefore defined as a revocation with no reinstatement in the following twelve months, with voluntary surrender separated out as a competing risk. Features freeze at month zero and the outcome is measured over months 6–18 — a deliberate blackout, because insurance lapse triggers revocation about a month later and a near-term model would just rediscover FMCSA’s own paperwork instead of finding signal ahead of it.
Thirteen silent data defects turned up along the way — a grand total read as a fleet size, a monthly export that was three files rather than one, a date format that quietly nulled an entire vintage. Each now has an automated guardrail and a regression test. That catalogue is written up in full in the repository, including the two defects that were in the leak-detection code itself.
Data ingestion through to a scored, ranked, decision-ready output
Which titles are worth buying ad inventory against? I scored all 500 titles in the 2025–26 global top 500 by blending how much they were actually watched with how long they stayed relevant. These charts run on the real output of that analysis — not screenshots.
Top Titles by Opportunity Score
Score blends hours viewed, sustained ranking, rating, and recency.
Hover or focus a bar for the underlying numbers. A bar that fades at its end runs past the axis — the axis stops above the runner-up so the rest stay comparable.
How the Score Is Weighted
Hours viewed dominates deliberately — it is the closest available proxy for realized ad impressions. The rest adjust for staying power and quality.
- Total hours viewed75%Strongest predictor of ad inventory value
- Weeks in Top 1013%Measures sustained cultural relevance
- Vote average8%Higher-rated titles hold attention longer
- Release year4%Newer content attracts more viewers
Both charts respond to these.
Genres by Average Score
Averaged across every ranked title carrying that genre. Title counts vary — War is a small, high-performing set.
Data notes and known limits
Genre coverage is partial: only 21 of the top 60titles carry genre metadata, because most 2025–26 chart-toppers are newer than the catalogue they join against. That is why there is no genre filter on the titles chart — it would silently hide two-thirds of the data. The genre panel aggregates across all500 ranked titles, where coverage is high enough to average meaningfully.
KPop Demon Hunters is a genuine outlier at roughly four times the second-place score, which compresses everything below it. It is left in rather than trimmed, because the whole point of the metric is to surface exactly that kind of title.
Data generated 2026-07-26 from the project's processed output.
RepositoryAnalysis notebookPower BI report
Design, build, and deployment
Every figure on this page renders from typed data, not markup. The career history, the impact metrics and both dashboards read from files under src/data/, so correcting a job title or a model score is a one-line edit rather than a hunt through templates.
The charts are hand-rolled SVG and CSS. A charting library would have added tens of kilobytes of JavaScript to draw rectangles that are already a known size at build time, and would have made the colour and contrast decisions harder to control rather than easier.
Colours are checked for contrast and for colour-vision deficiency separation in both light and dark themes; animation is decoration that prefers-reduced-motion turns off, never the only way to read something. The whole site is static, built and deployed by GitHub Actions on every push.
Get in touch
Open to analytics leadership roles
Based in Reno, NV, working remotely across time zones. If you are building something in the 0-to-1 space and need someone to make the data legible, I would like to hear about it.