Expertise
What I can be handed on day one.
Four domains, the methods inside each, and the projects on this site where they did real work. No percentage bars — a number next to a skill is a claim nobody can check.
01
Generative AI
LLM systems that do work, not demos.
- RAG & vector search
- Agentic workflows
- Tool calling / orchestration
- Prompt + eval pipelines
- LangChain, OpenAI, Claude, Gemini
Used in
02
Machine Learning
Forecasts and scores that move P&L.
- Demand forecasting (LightGBM)
- Churn & CLV models
- Recommendation systems
- Segmentation / RFM
- Probabilistic intervals
Used in
03
Data Science
From raw events to a decision.
- Pipelines & warehousing
- Experimentation / A/B
- BI: Power BI, Tableau
- GA4 + BigQuery
- Feature engineering
Used in
04
AI Systems Engineering
Deployed, observable, cost-aware.
- FastAPI / Django services
- Next.js interfaces
- Postgres + pgvector, Redis
- Docker, GCP, AWS
- Realtime voice & telephony
Used in
How a system gets built
The same six layers every time, from the signal nobody has wired up yet to the interface someone opens on Monday morning.
01
Data
Find the signal nobody has wired up yet.
Vessel movements, Shopify order streams, POS tickets, HR records, drawings and BoQ documents. Ingestion is scrapers, APIs, webhooks and file parsers — with verification and confidence levels attached at the point of capture, not bolted on later.
02
Processing
Make it trustworthy before it is clever.
Cleaning, entity resolution, deduplication and feature engineering. This is where confidence scoring, unit normalisation and leak-free train/serve parity get decided — most model failures I have debugged were really failures here.
03
Machine Learning
Predict the number someone will act on.
Gradient boosting for demand and churn, clustering for segmentation, regression for cost and CLV. Forecasts ship with intervals, because a planner needs the downside, not a single confident number.
04
LLM Reasoning
Language where rules are too brittle.
Retrieval over the business's own documents, structured extraction, tool calling and explanation. The model reasons; it does not invent facts — grounding and citation are part of the architecture, and evaluation runs on real examples.
05
Automation
Close the loop without a human in the middle.
Agents that act: send the outreach, recover the abandoned order, raise the reorder, book the appointment, escalate when confidence drops. Every action is logged, reversible and rate-limited.
06
Real-World Application
A product someone opens on Monday morning.
Interfaces that make the intelligence usable: dashboards, queues, chat, voice, reports. Deployed in Docker on GCP or AWS with observability and cost ceilings, because an AI feature that quietly burns tokens is a liability.
Tools I use in production
- AI & LLMs
- GPT
- Claude
- Gemini
- DeepSeek
- LLaMA
- Mistral
- LangChain
- RAG
- Hugging Face
- pgvector
- ML & Data Science
- Python
- R
- scikit-learn
- LightGBM
- XGBoost
- PyTorch
- TensorFlow
- NumPy
- Pandas
- SciPy
- Analytics & Visualisation
- Predictive modelling
- A/B testing
- Regression
- Clustering
- Power BI
- Tableau
- Matplotlib
- Altair
- GA4
- BigQuery
- Cloud & Data Platform
- GCP
- AWS
- Docker
- Apache Spark
- Hadoop
- Postgres
- MongoDB
- Redis
- DuckDB
- GitHub Actions
- Engineering
- FastAPI
- Django
- Node.js
- Next.js
- React
- TypeScript
- SQL
- Selenium
- Postman
- REST