People reporting you can trust
Dashboards built on one agreed definition per metric, so "turnover" means the same thing in every meeting.
- Headcount, attrition and org health
- Time to fill and quality of hire
- Compensation, probation and onboarding
HR analytics · Hiring automation · AI for talent teams
People data you can put in front of the board.
I help HR, talent and finance teams connect their systems, fix the numbers that don't add up, and automate the manual work that slows hiring down. That includes AI tools like interview briefings that pull from the systems you already use.
Engineering headcount, Q3
Example- HRIS export 212
- Finance plan 205
- ATS, marked hired 219
One number everyone usesRebuilt nightly from point-in-time history, with tests that flag drift.
212276changes shipped and reviewed in five months
206hiring-plan changes synced to the ATS in one validated upload
20+people and finance dashboards rebuilt or built new
30+data-quality bugs found and fixed at the source
Services
What I can take off your plate
Most HR teams already have the data. It's spread across an HRIS, an ATS, payroll and a finance model that each tell a slightly different story.
Hiring workflow automation
Request, approval and notification flows that keep your ATS, HRIS and finance plan in step.
- New role, backfill and change requests
- Bulk updates for planning cycles
- Reminders routed to the right manager
AI for talent teams
Practical tools that run on your own data and save interviewers and recruiters real time.
- Interview briefings sent before every call
- Draft scorecards after the interview
- Claude skills for your team's recurring work
Data foundation and ongoing support
The pipelines, models and tests underneath it all, plus an analyst on call when a number looks off.
- Warehouse, dbt models and orchestration
- Tests, freshness checks and alerts
- Monthly cost and usage breakdowns
Selected work
One engagement, five months
An embedded engagement with the People Analytics team at a multi-country tech scale-up. The client's name is withheld.
ClientTech scale-up, several countries
TeamPeople Analytics, TA, FP&A
RoleAnalytics engineer, embedded
ReportingWritten summary every month
The problem
The hiring plan lived in the ATS, the HRIS and the finance forecast, and the three drifted apart. Finance worried about roles being counted twice and about level or title changes made without approval.
What I built
- An internal app where hiring managers request new roles, backfills and changes, and approvers sign off
- A bulk-upload engine that validates large planning changes before they reach the ATS
- A change log showing the budget impact of every change, converted to the right currency and split by approval gate
- Automatic emails for backfill and out-of-plan requests, routed by role and org
~15 users in production206 reforecast changes in one upload129 open roles enriched with job family
The problem
Interviewers often joined calls without reading the application, and scorecards came in late or thin. That hurts candidate experience and makes quality of hire hard to measure.
What I built
- A briefing for each interviewer that combines the candidate's application, the role's scorecard and earlier feedback, sent in Slack before the call
- Automatic re-sends when an interview is rescheduled
- A draft scorecard after each interview, with a direct link to the ATS scorecard
- Coverage trackers showing which interviews got a briefing and a completed scorecard
- Two Claude skills the team uses for recurring work
Runs daily on the client's own warehouseAI compute reported separately each month
The problem
Report logic lived inside legacy BI views that nobody fully trusted, and the old BI environment was being shut down at the end of the quarter.
What I built
- Moved all BI view logic into tested dbt models
- Rebuilt 10 finance reports, including headcount and cost by business group and location, attrition and hiring forecast
- Rebuilt 5 people operations reports, including position data errors and time off
- New dashboards for performance, candidate experience, onboarding, org health, office attendance, absences and rewards
- New metrics: quarterly and rolling attrition, probation outcomes, Day 30 and Day 90 onboarding check-ins
20+ dashboards and reports10 stale models retiredMonth, quarter and year comparisons
The problem
Slow full refreshes of recruiting data, few tests, and warehouse costs that rose without a clear explanation.
What I built
- Moved recruiting data ingestion to Airflow with incremental loads, which is faster and cheaper
- Daily refresh jobs with alerts in a monitoring Slack channel
- Upgraded the dbt project and added tests and freshness checks across marts
- A cost breakdown by project, so leadership saw why spend rose 63% in a month before anyone had to ask
Along the way
Numbers that were wrong, and what fixed them
| Area | What was happening | Before | After |
|---|---|---|---|
| Interview counts | Counting logic multiplied interviews in the TA dashboards | 1,331 shown | 176 actual |
| Office history | Office came from each person's current record instead of their history, so past months were mislabeled after transfers | 269 employee-months wrong | Point-in-time history |
| Position IDs | A precedence bug tied applications to the wrong position | 151 applications | Fixed at the source |
| Job levels | An opening's level silently overrode the real offer level, showing a manager as an individual contributor | Wrong level | Offer and HRIS level win |
| Performance ratings | Ratings stopped updating after post-review calibration | Stale ratings | Calibrations synced |
| Hiring plan | Re-hire rows duplicated roles that were already filled, caught before a weekly business review | Double-counted roles | One row per role |
How I work
Small, reviewed changes on the tools you already pay for
Start with the question
What decision is this number for, and who needs it? The tool comes after that.
Build on your stack
Your warehouse, your BI tool, your ATS and HRIS. Nothing that stops working when I leave.
Ship in small pieces
Every change goes through code review and tests, so you can see exactly what moved and why.
Report in writing
A short summary every month of what shipped, what broke and what's next.
SnowflakedbtAirflowLookerStreamlitSnowflake CortexClaude CodeAshbyGreenhousePersonioCulture AmpSlackGitHub
Tired of rebuilding the same headcount report?
Tell me what your team is stuck on. I'll reply with how I'd approach it.