Engineering Professional

Sameera Bandarage

I design, build and run the data platforms that businesses depend on: high-throughput pipelines, well-modelled warehouses, and the trusted numbers that reports and AI systems are built on. I'm a certified engineer, and every tutorial here comes from a working lab I built myself on AWS, so you see real code and real results. Alongside data, I write about power electronics for renewable energy.

  • AI Systems Engineering
  • Data Engineering
  • Data Warehousing
  • Analytics Engineering
  • Data Architecture
  • Business Intelligence
  • Statistics
  • Power Electronics
Explore the tracks → See my expertise
Sameera Bandarage
Now writingAI Systems Engineering
Contactsameerabandarageln@gmail.com
8learning tracks
9tutorials
90published chapters
1working lab behind every example
  • WarehouseSQL Server 2022 · Snowflake · PostgreSQL
  • Transformdbt Core · Jinja · T‑SQL
  • OrchestrateApache Airflow
  • Model & BIPower BI · DAX · TMDL
  • AI systemsLLMs · RAG · vector search · agents
  • PlatformLinux · Docker · Kubernetes (EKS)
  • AnalysisPython · statistical inference
  • Electronics for Power EngineeringSiC · GaN · DC-DC · inverters

The platform, end to end

How the tracks fit together

The data and AI tracks each cover one layer of a modern data platform, following the data from the source system to the people and AI systems that use it. Select a layer to open its track.

01 · SOURCE Operational systems OLTP databases such as AdventureWorks, profiled and mapped before anything is loaded.
02 · LOAD & MODEL Data warehouse Extract to staging with watermarks, then load a star schema: SCD2 dimensions and incremental facts. Data Warehousing → 03 · TRANSFORM Transformation as code dbt models, tests and macros that build tested tables inside the warehouse. dbt → 04 · SERVE Semantic model One Power BI model with trusted measures, a fiscal calendar and row-level security. Business Intelligence →
05 · USE People & AI Analysistests, intervals, regressionStatistics → AI assistantsanswers grounded in the semantic modelAI Systems Engineering →
Orchestration · Airflow schedules and monitors steps 02–03, and the AI refresh jobs: retries, quality gates and re-runs Airflow chapters → Architecture · the design around all of it: reference ELT, Kubernetes and Linux VM deployments Data Architecture Design →

Power Electronics is a separate engineering track, outside the data platform.

Learning tracks

Eight tracks, one engineering toolkit

Every track goes past "getting started": how the technology works internally, why it behaves the way it does, and how to use it in real systems.

04Cloud platform · End to end

Snowflake

A complete platform on Snowflake: incremental loads from SQL Server, dbt Silver, Gold and a Sales data mart, reconciled and run by Airflow.

  • Snowflake
  • dbt
  • Medallion
  • SCD2
  • Key-pair auth
  • Airflow
06Beginner to advanced

Statistics

Practical statistics for everyday decisions, taught entirely with real oil and energy data.

  • Inference
  • Regression
  • ANOVA
  • Bootstrap
  • Bayes

Expertise

What I work on, and where I show it

Each area links to the chapters where I build it step by step, so you can check the work, not just read a list of skills.

Data modelling & warehousing

  • Kimball star schemas and grain design
  • SCD Type 1 and Type 2 dimensions
  • Incremental, idempotent fact loads
  • Columnstore, indexed views, Resource Governor

Shown in Dimensional Modelling · Loading Dimensions · Performance

Analytics engineering

  • dbt Core internals: parser, DAG, execution
  • Advanced Jinja and macro design
  • Dispatch and overriding dbt internals
  • Custom materializations and tests

Shown in dbt Core Internals · Custom Materializations · Capstone

Orchestration & data quality

  • Airflow DAGs with task groups and trigger rules
  • Watermarks that only move on success
  • Data quality gates and reconciliation
  • Running, failing and re-running safely

Shown in Orchestration · Data Quality Gate · Run, Fail & Re-run

BI & semantic modelling

  • Power BI star schemas and layered measures
  • 4-4-5 fiscal calendar and time intelligence
  • Currency conversion and row-level security
  • The model as code: PBIP, TMDL and git

Shown in Measure Architecture · Advanced Patterns · Model as Code

Data architecture & platforms

  • ELT reference architecture on Snowflake
  • Airflow on Kubernetes (EKS) and on Linux VMs
  • CI/CD, security, networking and sizing
  • Linux and Docker lab infrastructure

Shown in Reference Architecture · EKS Deployment · The Lab

Statistics & analysis

  • Confidence intervals and hypothesis tests
  • Regression, correlation and ANOVA
  • Bootstrap, permutation tests and Bayesian updating
  • Reproducible analysis on open data

Shown in Hypothesis Testing · Regression · Test Chooser

Cloud data platforms on Snowflake

  • Warehouses, databases and least-privilege roles, with a key-pair service user
  • Incremental loads: high-water marks, PUT to a stage and COPY INTO
  • Medallion layers in dbt: staging, SCD2 snapshots, conformed star schemas and a data mart
  • Tests and source reconciliation to the cent, run nightly by Airflow

Shown in Setting Up Snowflake · Extract & Load · Sales Data Mart · Reconciliation

AI systems engineering

  • LLM applications grounded in governed warehouse data
  • Embeddings, vector search and retrieval-augmented generation
  • Text-to-SQL that respects the semantic layer
  • Tool use and agents, evaluation, guardrails and LLMOps

Shown in the new AI & AI Systems Engineering track: chapters are being written.

Power electronics for renewable energy

  • Power semiconductors: Si, IGBT, SiC and GaN, and their loss models
  • PWM, switching transitions, dead time and thermal design
  • Buck, boost and buck-boost converter design for PV
  • Inverters, isolated converters, gate drivers and magnetics

Shown in Semiconductor Switches · Switching, PWM & Losses · DC-DC Converters

Built on a real lab

Every example runs, nothing is a mock-up

The code, numbers and screenshots come from a lab I run on a virtual machine in AWS: real loads, real failures and real re-runs.

Ubuntu + DockerOne Linux VM deployed in AWS, running every service in containers.
SQL Server 2022 on LinuxAdventureWorks sources and the SBEcoLTDDW warehouse.
Apache AirflowA 26-task DAG where each task is one SQL file.
dbt Core + PostgreSQLThe project behind the dbt tutorials.
Power BIA P&L model on AdventureWorksDW2022, stored as TMDL.
Open energy dataOil prices, refinery runs and Norwegian wells for statistics.

Where to start

Pick the path that fits you

Each track stands on its own, so you can start wherever your work is.

New to data warehousing Build a warehouse from scratch Start with the infrastructure and the big picture, then load it chapter by chapter. Start chapter 1 → Already using dbt See what happens inside dbt run Learn the engine, then move on to advanced Jinja and metaprogramming. Start chapter 1 → Power BI developer Model measures people can trust From star schema to measure architecture, tests and performance. Start chapter 1 →