Stephen Liu
Selected work

Operational savings, faster delivery, and AI systems people trust.

10 years across Alaska, Nordstrom, GE, and SchoolIntel — $9M+ in financial impact, 40% faster delivery, 73% fewer production issues, and 95%+ AI extraction accuracy.

10 years

experience across enterprise + founder roles

$9M+

documented financial impact

40% faster

deployment velocity

95%+

AI extraction accuracy

Trusted acrossAlaskaNordstromGESchoolIntel
Category index

Case 01 · Enterprise GenAI

RAG assistant with eval + trust guardrails

Story map
Permission boundaryConfluenceADOChatsMeetingsEvalharnessASSISTANTDecision answer+ source trace40%deployment velocity ↑
Case 01 / 03Alaska Airlines · 2024–2026
AI & GenAIProcess Excellence

Enterprise GenAI assistant for technical decisions

Principal Technical Program Manager

Turned scattered technical decisions into a secure RAG assistant teams could trust and actually use.

40%

faster deployment velocity

5 pods

coordinated across engineering

3 orgs

faster onboarding + decisions

Meeting transcriptsConfluence + ADO ingestionRetrieval evalsPermission-aware accessAdoption playbook
Read the full story
Context

A large airline IT org accumulated decisions across meetings, chats, and architecture docs faster than any team could absorb them. New hires lost weeks getting context; senior engineers re-litigated old debates.

Problem

Knowledge was high-volume, fragmented across systems, and politically sensitive. Off-the-shelf chatbots couldn't be trusted with the source material or the answers they produced.

Approach

Stood up a RAG-based assistant with curated ingestion (meeting transcripts, Confluence, ADO, chats), evaluation harness for retrieval and generation, and tight permissions. Drove platform decisions across cloud, security, and analytics. Designed adoption playbook so the assistant earned trust before broad rollout.

Complexity

Five engineering pods, security review, executive sponsors, and analytics dependencies. Required reframing GenAI as an evaluation problem, not a model problem.

Results
  • 40% increase in deployment velocity from intake to release
  • Faster onboarding and decision turnaround across analytics, engineering, and ITS
  • Recognized internally as Alaska Airlines AI Champion (2026)
02
next case · Data Quality

Case 02 · Data quality

Upstream anomaly detection before dashboards

Story map
UPSTREAM DOMAINSLOOPREDICRFLFIAnomaly detection gateshared correctness contractBLOCKED TRACK73%anomalies interceptedRELEASED TRACKTrusted datasetsto reporting + productsWEEKLY TRIAGE CADENCE
Case 02 / 03Alaska Airlines · 2024–2025
Data QualityAnalytics & BI

Data-quality product that earns stakeholder trust

Principal Technical Program Manager

Turned dashboard disputes into a monitored data-quality product that caught issues before they spread.

73%

anomalies caught pre-prod

95th %ile

stakeholder satisfaction

7 domains

aligned under one system

Anomaly detectionUpstream contractsWeekly triageExecutive reporting7-domain rollout
Read the full story
Context

Downstream products and executive reporting frequently disagreed on numbers. Each disagreement turned into a multi-week investigation; trust in data eroded across the org.

Problem

Data quality was a slogan, not a system. There was no shared definition of correctness, no anomaly detection, and no obvious owner for upstream issues.

Approach

Launched a data-quality product with anomaly detection at the upstream layer, paired with a clear contract for downstream consumers. Built operating cadences so teams triaged issues weekly instead of arguing in dashboards.

Complexity

Seven data domains, multiple business stakeholders, and an executive reporting layer that depended on the same pipelines.

Results
  • 73% of upstream anomalies caught before production
  • Stakeholder satisfaction in the 95th percentile
  • Automated executive reporting from Azure DevOps for VPs and managers
03
next case · AI & GenAI

Case 03 · SchoolIntel

9-agent verification graph + 19-step system

Story map
IGIngestEXExtractVRVerifySMSummarizeQAQASESEOPBPublishMNMonitorNTNotifyAUTOMATION RIBBON19 steps, automateddistrict research → ingestion → QA → SEO → publish1K+ MAU
Case 03 / 03SchoolIntel · 2025–2026
AI & GenAICustomer Experience

Multi-agent platform for parent-facing education intel

Founder / CEO

Built an agentic research platform that turns fragmented school data into parent-ready decisions.

1K+

monthly active users

95%+

extraction accuracy

19-step

automated onboarding system

9-agent pipelineVerification guardrails19-step onboardingSEO publishingEnd-to-end build
Read the full story
Context

Parents trying to evaluate school districts face fragmented board minutes, district sites, and ratings systems. The signal exists; the experience is missing.

Problem

Turning long, unstructured public documents into trustworthy parent-ready insights at the scale of thousands of districts — without hallucinations.

Approach

Designed a 9-agent AI pipeline with verification guardrails, plus a 19-step onboarding system that automated district research, ingestion, QA, SEO, and publishing. Built the product, the platform, and the operating system end-to-end.

Complexity

Solo founder operating across product, engineering, AI, content, and growth. Trust had to be built through accuracy, not marketing.

Results
  • Grew to 1K+ monthly active users in launch year
  • >95% extraction accuracy on board-document pipelines
  • Repeatable onboarding system that scales without proportional headcount
More case studies

Five more programs.

Compact accounts of process, FinOps, BI, and CX work delivered across Alaska, Nordstrom, and GE.

Hero stat
$3M

saved · 50% gate staffing reduction

Automated seat assignment strategy

Led automated seat assignment strategy and delivery, reducing gate staffing by 50%.

Alaska Airlines · 2023–2024
Categories
Process ExcellenceFinOps
Hero stat
VP cadence

Director + VP reporting on autopilot

Executive reporting automation across ADO

Automated executive reporting from Azure DevOps for VPs and managers across Analytics, Business, and ITS.

Alaska Airlines · 2024–2026
Categories
Analytics & BIProcess Excellence
Hero stat
+7 NPS

+10% adoption · 32% efficiency lift

Customer experience roadmap at scale

Drove alignment across 25+ teams through prioritization frameworks and standardized workflows, improving operating efficiency by 32%.

Nordstrom · 2021–2023
Categories
Customer ExperienceProcess Excellence
Hero stat
$3.2M

managed · 90% forecast accuracy

Multi-budget forecasting & global delivery

Managed $3.2M across four global budgets and improved forecast accuracy by 90% through Tableau-driven financial visibility.

GE Grid Solutions · 2020–2021
Categories
FinOpsProcess Excellence
Hero stat
+10%

patient treatment rates

Patient cancellation KPI system

Built seasonal and weekly cancellation-rate KPIs in Sisense and Power BI to drive treatment-rate decisions.

GE Healthcare Digital · 2018–2020
Categories
Analytics & BICustomer Experience
Résumé

A decade of programs.

View full CV →
Career timeline · 2016 → 2026
2016
2018
2020
2022
2024
2026
  1. 2025 — Present
    Founder / CEO
    SchoolIntelSchoolIntel

    Parent-facing education intelligence platform with a 9-agent AI pipeline and a 19-step automated onboarding system.

    Read case →
  2. 2024 — Present
    Principal Technical Program Manager
    Alaska AirlinesAlaska Airlines

    Enterprise RAG-based GenAI chatbot, data-quality product, and end-to-end CI/CD intake strategy.

    Alaska Airlines AI Champion (2026)Alaska Airlines Spot Bonus (2024)
    Read case →
  3. 2023 — 2024
    Manager, Business Intelligence (Mgmt Consultant)
    Alaska AirlinesAlaska Airlines

    Led a 7-person analyst team to 95th percentile stakeholder satisfaction; drove $3M seat-assignment savings.

    Alaska Airlines Spot Bonus (2023)
    Read case →
  4. 2021 — 2023
    Staff Technical Program Manager
    NORDSTROMNordstrom

    Aligned 25+ teams on a CX roadmap that improved NPS by 7 and adoption by 10%.

    Nordstrom Spot Bonus (2022)
    Read case →
  5. 2020 — 2021
    Senior Technical Program Manager
    GE VernovaGE Grid Solutions · GE Vernova

    Managed $3.2M across four global budgets; delivered three products totaling $2.7M on schedule.

    GE Above and Beyond Award (2020)
    Read case →
  6. 2018 — 2020
    Senior Business Intelligence Engineer
    GE HealthcareGE Healthcare Digital

    Built KPI systems that lifted patient treatment rates by 10%; led US financial reporting on AWS.

    GE Above and Beyond Award (2019)
    Read case →
  7. 2016 — 2018
    Technical Roles
    GE PowerGE Digital Technology Leadership Program · GE Power

    Automated $4M budget review (120+ hrs/qtr saved); integrated a $100M product line into S&OP planning.

    GE Above and Beyond Award (2017)
Education
  • M.S. — 2018
    Information Management
    University of Washington · Seattle
    Specialization: Program Management & Business Intelligence
  • B.S. — 2016
    Business Administration
    University of Pittsburgh · Pittsburgh
    Information Systems · Supply Chain Management
Certifications
  • Certified ScrumMaster (CSM)
  • Six Sigma Green Belt
  • Tableau
  • Google Project Management
  • Aha Product
Recognition
  • Alaska Airlines AI Champion (2026)
  • Alaska Airlines Spot Bonus (2023, 2024)
  • GE Above and Beyond Award (×3)
  • Nordstrom Spot Bonus (2022)
Next chapter

The work I do best sits between strategy, technology, and execution.

If your team is looking for a leader who can bring structure to ambiguity, align technical and business teams, and turn complex work into repeatable systems, I would be glad to talk.

Tell me about it →