Skip to content
View velurex's full-sized avatar

Block or report velurex

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
velurex/README.md

Lukáš Velecký

Senior .NET & Backend Engineer — building robust enterprise backends, cloud-native architectures, and distributed systems.

20+ years in software engineering with a deep foundation in systems and infrastructure. Focused on scalable C# / .NET backend services, mission-critical logistics and enterprise platforms, asynchronous job processing, and high-performance data architectures.


🚀 Currently working on

High-performance, memory-efficient inference for 1.58-bit ternary LLMs with 5 trits/byte encoding and division-free SIMD kernels.

1.58-bit LLMs SIMD Kernels Ternary Arithmetic Inference Optimization

💡 I’ve always been fascinated by optimization algorithms and bit-level tricks, and by how much can be gained by rethinking the arithmetic itself. That curiosity led me to explore efficient ternary multiplication, and quintrit grew out of it. It sits well outside my day-to-day .NET work and comfort zone, which is exactly why I enjoy it.


🧠 Recently learned

Recently, my study and hobby projects have all revolved around AI, developing across three distinct dimensions:

  1. Implementing AI & Agent Workflows — Integrating generative AI into real-world applications and mastering the surrounding ecosystem:
    RAG MCP Embeddings Vector DB Tool Calling ReAct Prompt Engineering Fine-tuning

  2. Production Guardrails & Reliability — Taming non-deterministic models and keeping AI within predictable bounds: continuous behavior monitoring, automated output validation, and strict guardrails and rules to ensure production-grade trustworthiness.

  3. LLM Internals & Architecture — Looking under the hood to understand the inner workings, layer structures, and low-level arithmetic of modern language models — the very insights that enabled me to design and build quintrit.


🤝 Happy to join forces on

I’d love to join a team or initiative focused on practical AI implementation — using AI to optimize domain processes and workflows. Beyond AI, I’m drawn to fields like energy and waste processing. These are new territories for me, and that’s the point: I want to step out of the .NET world and learn something new.

🎯 The projects I’m looking for should have broad, real-world relevance and societal benefit, while still having solid commercial potential.


💻 What I work on

  • 🏗️ Enterprise & Backend Systems — Scalable microservices and REST APIs in modern .NET (.NET 10/8), background job orchestration (Hangfire), clean architecture, and MediatR
  • ☁️ Cloud & DevOps — Cloud-native architectures on Microsoft Azure (App Services, Functions, Service Bus), Infrastructure as Code with Terraform & Bicep, CI/CD pipelines in Azure DevOps, Docker & Kubernetes
  • 🗄️ Data Engineering & Persistence — High-performance database design with MS SQL Server, Entity Framework Core, and Dapper for low-latency queries
  • ⚡ Full-Stack & Real-Time — Modern Blazor web applications, real-time messaging with SignalR, responsive web interfaces, and desktop applications (WPF/MVVM)
  • 📦 Logistics & High-Throughput Services — Mission-critical backend infrastructure powering real-time warehouse logistics, handheld terminal APIs, and enterprise ERP/audit workflows

🏢 Areas of Experience & Project Types

  • 🏭 Warehouse Logistics & Supply Chain — High-availability APIs and asynchronous background task pipelines supporting warehouse management systems and handheld scanner devices in fulfilment centers.
  • 🌐 Cloud-Native Comparison & SaaS Platforms — End-to-end architecture on Microsoft Azure, including serverless event-driven processing, real-time dashboards, and automated infrastructure deployments (IaC).
  • 🏛️ Enterprise & Public Sector Information Systems — Secure, auditable data management platforms, granular authorization, and data processing workflows handling complex business logic.
  • 🔐 Access Control & Facility Authorization Systems — Specialized client-server solutions for facility-wide asset authorization and tracking using WPF and REST APIs.

🛠️ Skills & Technologies

🖥️ Backend & Architecture

.NET C# ASP.NET Entity Framework Dapper MediatR SignalR REST API Unit Testing WCF

☁️ Cloud & DevOps

Microsoft Azure Azure Functions Azure Service Bus Azure DevOps Docker CI/CD Bicep GitHub Actions

🤖 AI & Systems

LLMs RAG MCP Embeddings Prompt Engineering Agent Workflows CUDA

💾 Databases & Persistence

MS SQL Azure Cosmos DB

🎨 Frontend & Desktop

WPF UWP Bootstrap CSS3 jQuery DevExpress Telerik

Popular repositories Loading

  1. LogusCore LogusCore Public

    LogoScore is an open-source Kotlin SDK for Android providing a modular abstraction layer for hybrid speech-to-text. It features provider neutrality, dynamic routing between local and cloud inferenc…

    Kotlin

  2. BitNet BitNet Public

    Forked from microsoft/BitNet

    Official inference framework for 1-bit LLMs

    C++

  3. llama.cpp llama.cpp Public

    Forked from isHuangXin/llama.cpp

    LLM inference in C/C++

    C++

  4. velurex velurex Public