DATA ENGINEERING · ML · AI SYSTEMS
The data speaks.
I build what makes it heard.
> now building: event-time correctness on 1.6M scan events
Ten projects across data engineering, machine learning and AI — 8.5 million simulated events, and every number in them reproducible from a command.
The work, in four statesSame data.
Same data.
Four states.
01 / 04
RAW
1.6M events · 31,615 duplicates
Collect
It arrives as noise.
Duplicated, out of order, and confident about both.
Engineer
Order is built, not found.
Deduped, resequenced, reconciled — until two teams get one answer.
Model
A shape, and how sure we are.
A forecast without a range is just an opinion with a chart.
Decide
One number, defensible.
Everything above exists so this one can be argued for.
Then the decision makes new data.
COLLECT→ENGINEER→MODEL→DECIDE↻
Three to start with
ALL TEN →Data Engineering Data Science / ML QuickCommerce Demand Intelligence & Personalization Engine
A 10-minute delivery promise only works if each dark store holds the right stock — not so much that it spoils, not so little that orders are lost — and if promotions reach customers who will actually use them.PythonXGBoostscikit-learnPySpark→ 1.3M order lines simulatedAI Engineering Fine-tuning Gemma 3 4B for Text-to-SQL with QLoRA
Turn natural-language questions into SQL for a banking database, using a model small enough to fine-tune on a free Colab T4.Gemma 3PEFT / LoRAQLoRAHugging Face Transformers→ Gemma 3 4B fine-tuned on a free T4
By the numbers
- 10Projects shipped
- 8.5MEvents processed
- 3Clouds: Azure, AWS, GCP
- 226Tests standing between data and dashboard