Themis Valtinos

0

Software Engineer @ Fivetran

Work across data transformation, query execution, table formats and analytical infrastructure, while also exploring compilers, machine learning, quantum computing and app development.

Scroll
01

Projects

Selected GitHub projects.

Tinkering with databases, compilers and machine learning.

DBMSC

Partitioned Hash Join

A partitioned hash join implementation for an in-memory columnar database focused on scalable execution.

Databases Algorithms Performance
View repository
NLPPython

Question Answering Transformers

BERT-based span entity and relation prediction over Wikidata, ending in SPARQL query generation.

BERT Wikidata SPARQL
View repository
CompilersJava

MiniJava LLVM Compiler

A MiniJava-to-LLVM compiler with semantic analysis, following the visitor pattern for LLVM IR generation.

LLVM Type Checking Compiler Design
View repository
GameC

Kiki's Delivery Service

A raylib side-scrolling platformer using custom ADTs written in C.

C Data Structures Game Dev
View repository
02

Apps

Shipped apps.

iOS apps spanning games and productivity tools.

iOSLive

Xeri (ξερή)

The classic Greek card game for iPhone, available with multiplayer mode.

Swift Game Center UIKit
Open project page
iOSLive

Habits

A habit planner and heatmap tracker for weekly planning.

SwiftUI Widgets Productivity
Open project page
iOSLive

Diloti (δηλωτή)

A faithful take on the classic Greek card game for iPhone.

Swift Card Game Strategy
Open project page
03

Open Source

Contributions.

Data transformation, table formats and cross-warehouse interoperability.

DataPython

SQLMesh

Main contributor to SQLMesh, around query planning, model execution and SQLGlot-based AST parsing.

Python Data Engineering Open Source
View repository
Query EngineC++

DuckDB Iceberg

Iceberg-side DuckDB work around parallel writes, extended type support and schema evolution.

DuckDB Iceberg C++
View repository
05

Research

Quantum Neural Networks.

Qudit-based feature maps and classification.

PaperSPIE

The Gell-Mann feature map of qutrits

A qutrit feature map using Gell-Mann rotations for classification tasks.

Read
PaperarXiv

Classification of data with a qudit

A geometric approach to qudit-based machine learning and classification.

Read

⚛︎ Overview

The research mainly explores Gell-Mann operators, quantum kernels and stacked PQC layers for qudit-level quantum neural networks. The parameterized quantum circuits encode data through feature maps, while variational parameters are optimized classically. Implementations of the quantum circuits are available on GitHub.

Let's build something meaningful.

For collaboration, consulting or feedback, reach out directly.