Brendan MairAI, product & problem solving

Selected work

Projects I own end to end.

Each one starts with a real question and ends with something you can open: the problem, my role, what I did and the outcome, next to the artifact I built.

  1. 01  Relay · Payment operations

    Explain every failed payment, and route around a failing processor.

    Problem
    When a payment fails, operations teams can't see why, and a failing processor keeps taking traffic until someone notices.
    My role
    Product owner and builder: scope, metrics, decision records, build and measurement.
    What I did
    Built a two-processor payment platform with one timeline per payment, plain-language failure explanations, automatic failover, incidents, dashboards, signed webhooks and a Python SDK.
    Outcome
    Complete on synthetic data: in a scripted processor outage, failover recovered most of the payments a static setup lost (simulated; figures on the project page).
    Read the case study
    Fig. 1 — Mock-up of the Relay console. All payments and processors are synthetic.

  2. 02  Regional Sales Data · Excel, SQL, Tableau

    Which sales channels actually make money, region by region.

    Problem
    Revenue showed which channels sold the most, not which ones earned the most once costs were counted.
    My role
    Analyst, end to end: question, data, dashboard, recommendation.
    What I did
    Cleaned and joined the sales data in Excel and SQL, then built profitability views by channel and region in Tableau.
    Outcome
    [2–3 findings from the deck] — owner to confirm
    Read the case study
    regional-sales/ profitabilityTableau view
    Top channel[channel]
    Margin gap[n] pts

    Profit margin by channel All regions

    Channel A[n]%
    Channel B[n]%
    Channel C[n]%
    Channel D[n]%
    Highest marginAverage
    Fig. 2 — Illustrative shape only. Channel names and values come from the Tableau workbook: owner to confirm.

  3. 03  Congress and Voter Behavior · R

    Does social pressure get people to vote, and by how much?

    Problem
    Turnout campaigns use social pressure, but how much it changes behaviour is rarely measured against a control group.
    My role
    Analyst: research question, statistical analysis in R, presentation.
    What I did
    Compared turnout across treatment groups and a control, estimating each effect with a confidence interval.
    Outcome
    Largest effect: [group], +[n] pts turnout — owner to confirm
    Read the case study
    congress-voting/ analysis.RR output

    lm(turnout ~ treatment) # effect vs control

    Group 1+[n]
    Group 2+[n]
    Group 3+[n]
    Group 4+[n]
    Fig. 3 — Illustrative shape only. Groups and estimates come from the analysis: owner to confirm.