Skip to content
View KlajdiBeqiraj's full-sized avatar

Block or report KlajdiBeqiraj

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. Markdown is supported. 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
KlajdiBeqiraj/README.md

Klajdi Beqiraj

AI Lead @ MYWAI. Applied AI across certified welding, underwater inspection and industrial process control, from edge data capture to cloud inference.

Electronic engineering background (MSc, University of Genoa, 110/110 cum laude), nearly six years in applied AI, from a master's thesis on industrial anomaly detection at Tenova to leading the AI component of a publicly funded research consortium.

What I actually work on

Most of it lives inside publicly funded research projects and consortium codebases, so it isn't on GitHub. Short version:

  • NEPTUNE (RAISE Liguria, NextGenerationEU/PNRR). I coordinate the AI vertical of a six-partner consortium and designed and built the NEPTUNE 5D Twins Studio, the platform's 3D viewer: seabed reconstruction via 3D Gaussian Splatting, vehicle trajectories, and probabilistic estimation of environmental fields from sparse probe measurements using Gaussian Process Regression. Vue + Three.js front end, Python/FastAPI back end. Also NEMO, the conversational mission-planning agent.
  • WELDWAISE (PR FESR Liguria, Polo EASS): agentic AI for certified welding. I built the multimodal retrieval system over welding events: per-modality encoders, temporal alignment across heterogeneous sampling rates, indexed in Qdrant. I also worked on the welding assistant itself: a voice interface over a locally hosted LLM, so no process data leaves the plant, with a real-time ML pipeline over machine and IMU streams.

Publications

First author on two papers at ASME Turbo Expo, on learned representations of 3D turbomachinery geometries (autoencoders and PointNet replacing manual parameterisation):

  • Advantages of Machine Learning Methods in Aerodynamic Blade Optimization, ASME Turbo Expo 2023 · 10.1115/GT2023-102481
  • Rotor37 Aerodynamic Optimization: A Machine Learning Approach, ASME Turbo Expo 2022 · 10.1115/GT2022-83063

What's here

Small Python tools, two of them packaged and published to PyPI. Not my AI work, but a fair sample of how I write and release code.

Stack

Python PyTorch FastAPI LLM & agents (agno, Flowise) Qdrant Gaussian Splatting Three.js Docker Azure Pipelines / ACR / ACA MQTT OPC UA

Elsewhere

LinkedIn · Medium · klajdibeqiraj96@gmail.com

Pinned Loading

  1. py_remote_recorder py_remote_recorder Public

    Remote screen and audio recording over HTTP. FastAPI service, Ngrok support, multi-monitor. On PyPI.

    Python 1

  2. VirtualAssistent VirtualAssistent Public

    Voice virtual assistant built on OpenAI GPT-3, December 2022. Speech-to-text, completion, text-to-speech.

    Python 1

  3. pyswaggerapiwrap pyswaggerapiwrap Public

    Generate Python wrappers for any Swagger-documented API. Filter routes, call them as attributes. On PyPI.

    Python