Data · automation · AI — Perth, WA → worldwide · EN / ES
I build the data platforms, automated reporting and AI agents that turn messy business data into decisions you can act on — for teams in Australia and around the world.
Dashboards that run themselvesAutomationAI agentsPredictive maintenance
Scroll — watch 37,768 real rows become a decision
01 — Noise
Dates in three formats, duplicates, totals that never match. In one Perth business alone: 10+ years of history across 20+ source tables.
−80%manual processing after the rebuild
02 — Pipeline
Ingest, model, test, serve — Python, SQL, BigQuery, dbt, Databricks — without breaking the reports people already use.
50 TB+refactored · −25% compute
03 — Signal
Reporting that runs itself. This landscape is 196,817 sessions across 90 days and 8 channels — every bar built from real rows.
30 → 3days per national reporting cycle
04 — Translated
Social brings 17.4% of the traffic and 0.3% of the revenue. Direct and Referral bring 72.8%. Same rows — now a decision.
4.44%Referral conversion vs 0.83% organic
05 — Intelligence
Agents that read your data and do the work — the same architecture behind a multi-channel AI platform across web, WhatsApp, SMS, email and voice.
01 · Built by me — dashboards & automation
Eight screens from a working analytics product. Each answers one business question with real data — tap one to open it, then keep going.
Drag the screens · tap one to open it
02 · Built by me — website factory (KarrotsKode)
KarrotsKode designs and ships websites and online stores for small businesses — AI writes from a brief, a studio-made design system does the rest. Every screen below came out of it.
The flagship · design, build, ship
A full product: brief to live site with HTTPS, a gallery of ready-made designs, custom AI sites and online stores with cart and checkout. I designed the platform, the generator and the design system behind every page.
03 · Built by me — try it live
A full analytics-engineering pipeline — BigQuery, dbt, DuckDB, Evidence — on Google’s public e-commerce sample, with natural-language SQL on top. Twelve pages of KPIs, drilldowns and data-quality checks.
Open the room →Predictive-maintenance twinA 3D drive you can take apart: inject a bearing fault and see the vibration spectrum, health index and remaining useful life react — the problem my PhD works on.
Open the twin →04 · Proof
Numbers from my CV — open any entry to see the role, the stack and the scale behind it.
Ready Chef GoMINVU · Ministry of Housing, ChileMetLife (via TCS)Tata Consultancy ServicesX-AnalyticIntexa IngenieríaMunicipal Corporation of Macul
−80%
manual processing
Three integrated data and reporting solutions for Ready Chef Go, a Perth food business, used by 20+ people.
30→3
days per reporting cycle
National housing reporting for Chile’s Ministry of Housing and Urban Development.
−25%
compute on 50 TB+
BigQuery models and SQL refactored for MetLife Chile without breaking downstream reporting.
−70%
model retraining time
Predictive-maintenance ML workflow across ~20 mining assets, automated end to end on AWS.
“As a project manager, Nicolás showed deep knowledge of industry and market trends and an outstanding ability to adapt to change in an agile way. He was particularly good at spotting opportunities to improve services — which led to innovative solutions that significantly increased customer satisfaction and business performance.”
“Nicolás brought solid financial and accounting knowledge together with a high level of technical skill. His command of technology tools made our move to new information-management and reporting systems much easier.”
“Certificate of Appreciation — recognised as Golden Guru at the Golden Guru Gala 2023, for the time, effort, dedication, passion, patience, knowledge and emotion brought to every session.”
05 · Offers
Scoped, fixed-price stages — quoted after a short call. Start small, keep what works.
Any team that doesn’t trust its numbers
An audit of your sources, reports and manual steps — with quick wins and a costed roadmap.
1 week · fixed scope
See: “Can you trust these numbers?” →Finance and operations leads
Pipelines, modelled KPIs and dashboards on your stack, refreshed automatically — no more month-end rebuilds.
2–4 weeks
See: the Live Data Room →Teams buried in repetitive questions and tasks
An agent that reads your data and does the work — retrieval, tool calling and memory — on web, WhatsApp, email or voice.
3–4 weeks
Try: my AI twin →Mining, industry and energy
From sensor data to anomaly detection, remaining useful life and alerts your planners can act on.
6–8 weeks · scoped together
Try: the predictive-maintenance twin →Teams after the first win
Ongoing support: new reports, monitoring, model upkeep and the next automation — one engineer who already knows your data.
Monthly
See: results with receipts →SQL · Python · Power BI · BigQuery · Databricks · dbt · Spark · AWS · Azure · GCP · OpenAI · RAG · LangGraph · FastAPI · PyTorch
06 · About
Ten years across enterprise, government, mining and Australian businesses. I started in finance, operations and management control — which is why I build systems that answer business questions, not just run. Today I combine hands-on engineering with end-to-end ownership, and I’m doing a PhD in Computing at Curtin on machine learning for predictive maintenance and intelligent industrial systems.
07 · Questions
Yes. I’m based in Perth (AWST) and work with teams across Australia and overseas, in English or Spanish. We agree overlap hours for calls; the work itself runs asynchronously.
Usually a one-week Data Health Check or a small pilot with one measurable outcome. You see working results early, and decide the next step with real numbers.
Fixed scope per stage, quoted after a 20-minute call — so you know the cost before anything starts. Ongoing support is a monthly arrangement.
That’s the normal starting point. At Ready Chef Go I reconciled 10+ years of history across 20+ source tables before automating anything.
Yours first — Excel, SharePoint, Power BI, MYOB, BigQuery, Databricks, Fabric, AWS, Azure or GCP. Open-source where it saves you money, never lock-in for its own sake.
Yes. I work in your tenancy with least-privilege access, and nothing leaves it unless we agree. Happy to sign your NDA before we look at any data.
Through KarrotsKode, the platform I built — that’s where the 145 designs above come from. My own work for clients is data, automation and AI.