Project library

Full project depth, proof surfaces, and recruiter-readable outcomes.

The homepage now stays brief by design. This page carries the full project library, translation layer, and artifact vault so every major project can be treated seriously without overcrowding the landing page.

Browse projects Open web CV Contact Michael

Major projects

Equal-weight case studies spanning BI infrastructure, AI orchestration, predictive analytics, NLP, regression, and database design.

Proof policy

Public content stays derived and sanitized. Raw resumes, raw decks, and internal source material remain local-only inside the intake system.

Project library

Equal-weight project depth, without homepage bloat.

This is the canonical project surface for the portfolio. Every major project gets the same level of seriousness: problem, method, result, public-safe proof path, and recruiter-readable outcome.

Power BI + stakeholder adoption

Command Center BI Infrastructure

Live

sanitized

High confidence

Built scalable Power BI reporting systems that turned fragmented sales data into decision-ready operating visibility.

Headline outcome

Reporting cadence scaled beyond manual capacity

Power BIDAXPower QueryVBADatabricksStakeholder Discovery

Impact

20+ hours/week removed from recurring reporting and follow-up effort.

Proof path

• Sanitized BI infrastructure brief
• Request-only dashboard walkthrough
• Reporting template and automation notes

Problem

Stakeholders needed cleaner opportunity visibility, KPI logic, recurring reporting, and adoption tracking inside a growing internal analytics hub.

Method

Used Power BI, DAX, Power Query (M), VBA automation, stakeholder discovery, QA, and documentation to translate messy asks into durable reporting logic.

Result

Delivered a reusable reporting backbone for executive visibility, self-serve decision support, and cleaner adoption tracking.

Review sanitized brief

Agentic workflows + durable context

Gemini/Codex Workflow Automation

Live

public

High confidence

Built a file-driven agentic workflow system that keeps Gemini CLI and Codex CLI aligned through durable context, reconciliation, and validation gates.

Headline outcome

Multi-agent AI work stopped depending on chat memory

ShellPythonGemini CLICodex CLIMCPn8n

Impact

Working system with 100 scripts, 34 docs, and 58 research artifacts in the read-only source repo.

Proof path

• Sanitized workflow brief
• Public docs repository
• Architecture and runbook excerpts

Problem

Long-running AI coding and business-automation workflows break when context, handoff, permissions, and validation live only in chat memory.

Method

Designed a constitution-first automation environment around declarative intent, a reconciliation loop, blackboard coordination, Gemini/Codex bridge scripts, MCP and n8n lanes, and deterministic close-out gates.

Result

Created a working orchestration environment that treats AI collaboration like an operational workflow system instead of a one-off prompt chain.

Open proof brief Open docs repo

Adidas IT · Gradient Boosting

Ticket Reassignment Prediction

Completed

sanitized

High confidence

Modeled IT ticket reassignment behavior to reduce escalation friction, rerouting cost, and downtime risk.

Headline outcome

Preventable routing friction became a measurable prediction target

Pythonscikit-learnFeature EngineeringModel Evaluation

Impact

Approximately 76.1% accuracy, 85.8% recall, 73.3% F1, and about $277K in annual labor savings, with additional downtime savings cited in the project materials.

Proof path

• Sanitized methodology page
• Downloadable presentation deck
• Model evaluation summary

Problem

IT support workflows were losing time and retail uptime through preventable ticket reassignments.

Method

Built a Gradient Boosting classifier around urgency, service, category, and geography variables, then evaluated it with accuracy, recall, precision, and F1.

Result

Produced a predictive decision aid and business-facing routing recommendation that could flag likely reassignments earlier in the ticket flow.

Open methodology page

Regression + cross-validation

Spotify Popularity Prediction

Completed

public

Medium confidence

Used audio-feature data to model song popularity and translate statistical results into business-readable recommendations.

Headline outcome

Turned audio features into business-readable popularity signals

PythonRegressionCross ValidationFeature AnalysisEDA

Impact

Statistically significant feature analysis and model interpretation, positioned as an interpretable analytics case study rather than a hard business-impact claim.

Proof path

• Recorded walkthrough
• Sanitized regression brief

Problem

Needed to understand which measurable audio features aligned with song popularity rather than relying on intuition.

Method

Applied regression, correlation analysis, exploratory data analysis, outlier handling, and cross-validation-oriented evaluation to a Spotify feature dataset.

Result

Produced an interpretable modeling case study that showed which features mattered most and how to communicate them without overstating prediction.

Open regression brief

TF-IDF + logistic regression

Yelp Review Rating / Sentiment Modeling

Completed

public

Medium confidence

Built a text-mining workflow to classify Yelp review sentiment and surface operational signals from customer language.

Headline outcome

Customer review text became a usable service-improvement signal

PythonTF-IDFLogistic RegressionVADERFLAIR

Impact

High-accuracy sentiment classification paired with operational feedback a manager could act on, as documented in the final project materials.

Proof path

• Recorded walkthrough
• Downloadable presentation deck
• Methodology report

Problem

Needed to turn large volumes of review text into structured signals for customer sentiment and service issues.

Method

Applied text preprocessing, TF-IDF vectorization, VADER and FLAIR sentiment labeling, and logistic regression classification with standard evaluation metrics.

Result

Generated a model and business readout that translated review language into repeatable service-improvement signals.

Open NLP brief

Multivariable regression

TJIX Net Sales Drivers

Completed

public

High confidence

Linked advertising, e-commerce growth, and market trends to a business-readable net-sales growth story for TJX.

Headline outcome

Advertising and e-commerce were tied directly to net-sales upside

RegressionCorrelation AnalysisExcelDecision TreesScenario Modeling

Impact

Final materials attribute roughly $12.7M in net-sales lift per additional $1M in advertising spend.

Proof path

• Live methodology page
• Downloadable final report
• Downloadable workbook

Problem

Needed to explain which levers most credibly drive TJX net sales growth and where e-commerce and advertising investment mattered.

Method

Used regression, correlation analysis, scenario modeling, Excel, and competitor comparison to tie advertising and e-commerce growth to net-sales outcomes.

Result

Produced a business-facing recommendation for deeper e-commerce and advertising investment backed by regression and benchmark analysis.

Open report bundle

Access + SQL queries

Relational Database Design

Completed

public

High confidence

Designed a relational data model to replace spreadsheet-driven project and staffing tracking for a growing IT support business.

Headline outcome

Operations data became structured, relational, and report-ready

Microsoft AccessSQLERD DesignNormalizationMetadata Design

Impact

Replaced spreadsheet thinking with a structured data model built for queryability, reporting consistency, and future operational growth.

Proof path

• Live methodology page
• Downloadable ERD PDF

Problem

Spreadsheet-based staffing and project tracking was causing inconsistency, limited analysis, and poor scalability.

Method

Designed normalized entities, bridge tables, metadata rules, and query patterns for projects, employees, clients, assets, tickets, and services.

Result

Produced a relational structure built for cleaner reporting, resource visibility, and future operational growth.

Open database brief

Impact lab

Business leverage deserves its own space, not a cramped homepage corner.

This is where the quantified story lives: modeled savings, automation compression, and recruiter-readable signal around what the work can do for a team when it moves from manual effort into repeatable systems.

Impact Analysis Dashboard

Recruiter-friendly proof of scale, business value, and technical range.

live signal

IT Routing Efficiency

Estimated annual savings

$277,000

Python, Gradient Boosting

Manual Reporting Automation

Value of reclaimed analyst time

$52,000

VBA, Power Query, DAX

Sales Driver Identification

Revenue signal surfaced

$12,700,000

Regression, stakeholder analysis

The ROI of Automation

Based on my work at Avnet (converting a 5-hour manual process to a 30-minute script), calculate how much I could save your team.

Manual Reporting (Hours/Week)10 hrs

Average Team Hourly Rate ($)$50/hr

Time Reclaimed (Yearly)

442 hrs

Estimated Cost Savings (Yearly)

$22,100

Translation layer

Raw business signal scanned into recruiter-ready proof.

This strip turns the actual shape of the work into something a recruiter can process fast: messy asks, KPI logic, and workflow notes crossing a scanner line and resolving into proof assets.

Raw signal

Scroll to scan

Proof asset

Power BI / KPI translation

Executive visibility surface

Command Center BI proof brief

Fragmented opportunity reporting becomes a recruiter-ready walkthrough of KPI logic, adoption flow, and operating visibility.

Review sanitized brief

Raw stakeholder askSCAN-2026-PROOF-1001

// PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1001 // // STAKEH
OLDER_ASK: show pipeline clarity by region, owner, and
 segment // DEFINE KPI.win_rate = DIVIDE([wins],[oppor
tunities]) // GROUP BY fiscal_week, account_owner, mot
ion // REQUIRED_OUTPUT = executive trend view + self-s
erve drill path // OUTPUT.surface = Executive visibili
ty surface // OUTPUT.story = Command Center BI proof b
rief // // PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1001 // /
/ STAKEHOLDER_ASK: show pipeline clarity by region, ow
ner, and segment // DEFINE KPI.win_rate = DIVIDE([wins
],[opportunities]) // GROUP BY fiscal_week, account_ow
ner, motion // REQUIRED_OUTPUT = executive trend view 
+ self-serve drill path // OUTPUT.surface = Executive 
visibility surface // OUTPUT.story = Command Center BI
 proof brief // // PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1
001 // // STAKEHOLDER_ASK: show pipeline clarity by re
gion, owner, and segment // DEFINE KPI.win_rate = DIVI
DE([wins],[opportunities]) // GROUP BY fiscal_week, ac

Durable context / AI orchestration

Public systems proof

Gemini/Codex workflow proof brief

Bridge scripts, blackboard coordination, and reconciliation logic resolve into a public-safe proof brief with a live docs-repo handoff.

Open proof brief

Agent orchestration contractSCAN-2026-PROOF-1002

// PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1002 // DECLARE d
esired_state = INTENT.md // LEASE blackboard.owner = a
tomic_file_lock // ROUTE bridge = gemini_cli <-> codex
_cli // CLOSEOUT = sync + check + archive + push // OU
TPUT.surface = Public systems proof // OUTPUT.story = 
Gemini/Codex workflow proof brief // // PORTFOLIO_SIGN
AL: SCAN-2026-PROOF-1002 // DECLARE desired_state = IN
TENT.md // LEASE blackboard.owner = atomic_file_lock /
/ ROUTE bridge = gemini_cli <-> codex_cli // CLOSEOUT 
= sync + check + archive + push // OUTPUT.surface = Pu
blic systems proof // OUTPUT.story = Gemini/Codex work
flow proof brief // // PORTFOLIO_SIGNAL: SCAN-2026-PRO
OF-1002 // DECLARE desired_state = INTENT.md // LEASE 
blackboard.owner = atomic_file_lock // ROUTE bridge = 
gemini_cli <-> codex_cli // CLOSEOUT = sync + check + 
archive + push // OUTPUT.surface = Public systems proo
f // OUTPUT.story = Gemini/Codex workflow proof brief 
// // PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1002 // DECLAR

Regression / downloadable proof

Download-ready bundle

TJIX report and workbook bundle

Regression notes, scenario framing, and workbook logic resolve into a downloadable proof bundle that keeps the business story intact.

Open report bundle

Commercial model handoffSCAN-2026-PROOF-1003

// PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1003 // SOURCE = 
ad_spend, ecommerce_penetration, competitor_growth // 
FIT regression = revenue_signal_story // EXPORT report
 = recruiter_safe_pdf // EXPORT workbook = scenario_mo
del_bundle // OUTPUT.surface = Download-ready bundle /
/ OUTPUT.story = TJIX report and workbook bundle // //
 PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1003 // SOURCE = ad
_spend, ecommerce_penetration, competitor_growth // FI
T regression = revenue_signal_story // EXPORT report =
 recruiter_safe_pdf // EXPORT workbook = scenario_mode
l_bundle // OUTPUT.surface = Download-ready bundle // 
OUTPUT.story = TJIX report and workbook bundle // // P
ORTFOLIO_SIGNAL: SCAN-2026-PROOF-1003 // SOURCE = ad_s
pend, ecommerce_penetration, competitor_growth // FIT 
regression = revenue_signal_story // EXPORT report = r
ecruiter_safe_pdf // EXPORT workbook = scenario_model_
bundle // OUTPUT.surface = Download-ready bundle // OU
TPUT.story = TJIX report and workbook bundle // // POR

Predictive analytics for recruiters

Business-readable methodology

Ticket routing methodology brief

Model notes, evaluation metrics, and business framing resolve into a concise predictive-analytics proof surface with a live deck handoff.

Open methodology page

Analysis notesSCAN-2026-PROOF-1004

// PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1004 // MODEL = g
radient_boosting_classifier // METRICS = accuracy, rec
all, f1_score // QA = validation_split + stakeholder n
arrative alignment // DELIVERABLE = one_to_two_page_me
thodology_brief // OUTPUT.surface = Business-readable 
methodology // OUTPUT.story = Ticket routing methodolo
gy brief // // PORTFOLIO_SIGNAL: SCAN-2026-PROOF-1004 
// MODEL = gradient_boosting_classifier // METRICS = a
ccuracy, recall, f1_score // QA = validation_split + s
takeholder narrative alignment // DELIVERABLE = one_to
_two_page_methodology_brief // OUTPUT.surface = Busine
ss-readable methodology // OUTPUT.story = Ticket routi
ng methodology brief // // PORTFOLIO_SIGNAL: SCAN-2026
-PROOF-1004 // MODEL = gradient_boosting_classifier //
 METRICS = accuracy, recall, f1_score // QA = validati
on_split + stakeholder narrative alignment // DELIVERA
BLE = one_to_two_page_methodology_brief // OUTPUT.surf
ace = Business-readable methodology // OUTPUT.story =

Stakeholder asks, KPI definitions, SQL, DAX, and workflow notes feed the left side.

The center scan line visualizes Michael’s strongest skill: translation under ambiguity.

The right side resolves into recruiter-speed proof assets already wired into the artifact vault.

Artifact vault

Proof surfaces, downloads, and recruiter-safe artifacts.

This vault now mixes live proof pages, downloadable artifacts, and request-only recruiter walkthroughs. The most publishable project proof leads, and anything still planned stays clearly labeled as planned.

Request-only proof

Request-only

sanitized web brief

dashboard

Command Center BI proof brief

Projects · Translation layer · Artifact vault · Assistant

Sanitized recruiter-safe proof page for the Command Center BI work, including the problem, reporting method, measurable outcome, and the request-only path for a deeper walkthrough.

Live dashboards remain private, but this brief documents the BI infrastructure and the recruiter-safe disclosure path.

Review sanitized brief

Live proof

sanitized web brief

pdf

Gemini/Codex workflow proof brief

Projects · Artifact vault · Assistant

Public-safe brief covering durable context, reconciliation, bridge scripts, and validation gates in the Gemini/Codex workflow system.

Pairs a sanitized web brief with direct links to the public docs repo and architecture receipts.

Open proof brief

Live proof

methodology page + deck

pdf

Ticket routing methodology brief

Projects · Artifact vault · Assistant

Sanitized project page for the Adidas ticket reassignment model, with the downloadable deck and business-facing methodology path.

The live proof surface is the methodology page and downloadable presentation deck.

Open methodology page

Live proof

methodology page + downloads

template

TJIX report and workbook bundle

Projects · Artifact vault · Assistant

Live proof bundle for the TJIX net-sales analysis, including the methodology page, native report, and supporting workbook.

This proof surface includes both the commercial regression writeup and the supporting workbook.

Open report bundle

Live proof

methodology page + PDF

pdf

Relational database design brief

Projects · Artifact vault · Assistant

Project page covering the normalized database design, ERD logic, and the public-safe PDF walkthrough.

The live proof surface includes the methodology page and the downloadable ERD PDF.

Open database brief

Live proof

methodology page + walkthrough

video

Yelp sentiment modeling brief

Projects · Artifact vault · Assistant

Project page for the NLP classification workflow, with the recorded walkthrough plus downloadable deck and report.

The strongest public proof here is the recorded walkthrough paired with the downloadable project materials.

Open NLP brief

Live proof

methodology page + walkthrough

video

Spotify regression brief

Projects · Artifact vault · Assistant

Sanitized project page for the Spotify regression case study, focused on interpretable modeling and the recorded walkthrough.

The live proof surface is the recorded walkthrough. The full final report is not yet published publicly.

Open regression brief

Live repo

Live proof

public docs repo

repo

Gemini/Codex workflow docs repository

Projects · Artifact vault · Assistant

Live public documentation layer for the Gemini/Codex automation system, including architecture notes, runbooks, and readiness receipts.

Public-safe companion repo for a much larger local automation workspace and CLI orchestration system.

Open docs repo