Business analysis, systems, AI automation
BUSINESS.
BEYOND
ORDINARY.
Your business. On autopilot.
Business analyst and project manager. I build AI automation, data pipelines, and system architecture that run operations without you.
intelligence layerENTER THE SIGNAL LAB
Human ambition. Machine precision.
Systems in production, not slideware.
03 systemsin the real world
Hidden in
your data.
Change the question.
See a different business.
Find the signal for yourself.
Where is the momentum?
The trend is clearer than the noise.
Explore a fictional year of revenue. Change the target to see which months qualify, then inspect a month for its exact value.
Fictional data, real calculations. This is an interactive illustration, not a client dashboard.
View the sample data
| Month | Value |
|---|---|
| Jan | $24k |
| Feb | $28k |
| Mar | $25k |
| Apr | $39k |
| May | $36k |
| Jun | $48k |
| Jul | $44k |
| Aug | $62k |
| Sep | $56k |
| Oct | $74k |
| Nov | $69k |
| Dec | $88k |
Enterprise AI data agent and predictive analytics dashboard
Measured impact
- customer churn within two quarters
- −38%
- model AUC in production
- 0.91
- question to answer
- 3d → 45s
- 01executive questions
- 02governed SQL warehouse
- 03verified
- 04answers
The challenge
Churn appeared in the numbers sixty days after it started. Every deep-dive question meant an analyst, a SQL export, and a three-day wait.
The system
A conversational AI agent over a governed SQL warehouse answers executive questions with verified, read-only queries. A churn model rescoring nightly in Python feeds live dashboards: every at-risk account surfaces with its statistical drivers and a recommended play.
SQL / Python / scikit-learn / Claude API / PostgreSQL
Chapter 02 / 05CRM · automation architecture
Scroll to move through the story
Make it connect.
End-to-end CRM operations architect
Measured impact
- hours of manual work eliminated every week
- 40+
- first-call resolution
- +24%
- records re-keyed by hand
- 0
FROM MESSY INPUT TO
A CLEAN
HANDOFF.
Fictional records and agents. Processed entirely in your browser. Nothing is sent.
- 01NORMALIZE
- 02DEDUPLICATE
- 03ROUTE
INCOMING
5 events- R-104Web formMAYA REED[email protected]
- R-105Web formleo park [email protected]
- R-106Emailnora BELL[email protected]
- R-107CallMaya Reed[email protected]
- R-108Web formELI COLE [email protected]
Mixed casing. Extra spaces. One repeat contact.
ASSIGNED
00 / 04Run the sample. Watch each contact find its team.
- INCOMING
- 05
- MERGED
- —
- ASSIGNED
- 00
Awaiting your first run
How the sample assigns contacts
Matching emails merge into one contact, keeping both source references and the higher priority. The latest incoming record supplies the other fields. Each topic routes to a fixed sample team:
- BillingIris Hale
- OnboardingTheo Park
- SupportNina Vale
Sample ready. Five fictional records.
The challenge
A 60-agent call center re-keyed every interaction across three systems, routed callbacks by gut feeling, and burned shifts on reports nobody trusted.
The system
An event-driven layer around the CRM: webhooks sync every disposition in real time, bots run wrap-up and data hygiene autonomously, and a predictive router sends each callback to the agent most likely to close it first time.
Webhooks / REST APIs / n8n / Predictive routing / TypeScript
Chapter 03 / 05Finance · executive BI
Scroll to move through the story
Make it clear.
Advanced financial analytics and executive BI pipeline
Measured impact
- days to close monthly reporting
- 12 → 1
- reconciliation accuracy
- 99.8%
- version of the truth
- 1
- 01ERP, bank, and billing exports
- 02one governed metrics layer
- 03board-grade BI surface
The challenge
Month-end close meant five analysts reconciling ERP, bank, and billing exports by hand. The board pack arrived late, and no two versions of revenue ever matched.
The system
A statistical cleanup pipeline feeds one governed metrics layer. Executives get a board-grade BI surface where every metric traces back to validated transactions, built to be read in the room, not deciphered after it.
Python / pandas / dbt / SQL / Power BI
Pipeline / Hands on
Live sandbox. Run the nightly pipeline, or type to build it by hand.
dani@prod: ~/pipelines
$ ▍Press run to execute the nightly churn pipeline, or click in and type on any keys to build it by hand.
Every system ships through the same four gates.
No improvisation and no black boxes. Each phase closes with artifacts you keep, whether or not we continue.
- 01
Discovery and data auditing
Map where the operation loses time and money, extract the existing data, and put a number on the target ROI before any code is justified.
OutputTarget ROI - 02
Architecture and schema design
PostgreSQL and Supabase modeling, end-to-end flowcharts, and the prompt-routing topology that decides which model handles what, and what never touches a model.
OutputSchemas + flowcharts - 03
Core engineering and automation
Python pipelines, secure API and webhook integrations, automation bots, and LLM orchestration with validation gates. Version-controlled, reviewable, never held hostage.
OutputPipelines + validation - 04
Deployment and analytics guardrails
Production release with instrumentation from day one: live BI dashboards and KPI guardrails that alert when reality drifts. The system proves its ROI instead of asserting it.
OutputKPI guardrails
Chapter 04 / 05Systems that work while you sleep.
Scroll to move through the story
What’s possible next?
I translate operational chaos into systems that measure, automate, and improve themselves.
I work at the intersection of business and engineering. As a business analyst and technology project manager, I have led digital transformation end to end: stakeholder discovery, process mapping, architecture, vendor selection, delivery, adoption, and measurable ROI.
My rule for every engagement is simple. If a process runs on someone's memory, a spreadsheet, and hope, it can be turned into a system. And once it is a system, it can be measured, automated, and improved.
How I work
- 01
End-to-end ownership, from discovery through adoption
- 02
Fluent in boardroom language and technical specification alike
- 03
Data-first decisions: every recommendation backed by numbers
- 04
AI pragmatist: automation where it pays, humans where it matters
Tools
- SQL
- Python
- PostgreSQL
- Supabase
- Claude API
- n8n
- dbt
- Power BI
- Next.js
- TypeScript
- Webhooks
- CI/CD
- Years in tech, data and operations
- 0+
- Systems and workflows automated
- 0+
- Manual hours eliminated yearly
- 0K+
- Cross-functional projects delivered
- 0+
Chapter 05 / 05A conversation. A clearer direction.
Scroll to move through the story
Your next big move.
LET’S MAKEIT WORK.
Fifteen minutes. One process. A concrete first move.
Not a discovery call, and nobody will try to sell you anything. Bring your messiest process; we map it live, locate the automation seams, and you leave knowing what to do next.
Book an auditRead transcript
If your team spends too much time on reports or manual tasks, Dani can help. Leave your details below, and he’ll get back to you to discuss it.
LOOK
through.
What looks like business as usual
has a whole other story inside.Move the lens. Follow a connection.
Find the work that shouldn’t be there.
Every connection tells a story.
0 / 3 SIMULATED FIXESCaptured. But never claimed.
Three enquiries reached the CRM. None has an owner or a next action. The record exists; the responsibility does not.
Assign an owner on arrival, attach a follow-up, and escalate anything left unclaimed.
| RECORD | SYSTEM | OBSERVATION | STATE |
|---|---|---|---|
| L-104 | Web form → CRM | Captured at 09:12 | Unassigned |
| L-108 | Web form → CRM | Captured at 09:21 | Unassigned |
| L-112 | Web form → CRM | Captured at 09:39 | Unassigned |
No real systems are modified.
15-minute system architecture audit
A direct working session with me. One process, mapped and diagnosed live.
- 01
A systems map of your highest-friction workflow
- 02
The two or three automation seams with fastest payback
- 03
A plain verdict: automate, redesign, or leave alone
3 audit slots per week
Send an inquiry
Describe your challenge. I reply within one business day.
Prefer email?
[email protected]