Victor Luncașu

• Field of study: Artificial Intelligence / Machine Learning

• Thesis: Creating a Hosting Service with OpenCore-run System Automation — graded 10/10
• Field: Engineering, manufacturing and construction

• End-to-end development of mobile (iOS/Android), web, and native desktop applications
• Architected and operates a distributed-compute and hosting platform
• On-premise AI solutions with proprietary local computing engines
• Manages and mentors a team of 10 developers and DevOps engineers

• Online and on-site store for high-performance workstations and server solutions
• Custom enterprise-grade configurations deployed on Proxmox virtualization

• Automated embedded-hardware validation frameworks — reduced manual testing by 40%
• CI/CD pipelines (Jenkins, GitLab CI) for firmware builds and deployment
• Modular libraries for device communication (UART, I²C, SPI) and log parsing

• Deployed regional IT network (Bucharest & Transylvania): VPN tunnels, AD integration
• Automated VM provisioning and patch-management scripts

Photorealistic aircraft configuration platform built in Unreal Engine → migrated to Babylon.js/WebGL. Configurable exterior liveries, propulsion systems, avionics, and interiors with compatibility/pricing logic, automated server-side rendering, distributed GPU workers, render-job orchestration, and caching infrastructure.

Rotating hydroponic hardware prototype — ESP32, pH/EC/temperature sensors, camera, stepper motor, pumps, LED lighting. Software: React Native, Node.js, Express, MQTT, PostgreSQL, closed-loop PI control. CNN for plant-condition classification; Random Forest for nutrient-effect modelling.

Commerce-ready platform for uploading 3D models, converting engineering files (STEP→STL), previewing parts in WebGL, estimating cost via adapted Bambu slicer CLI, and server-side Blender rendering. Distributed queues, load balancing, Docker on Proxmox.

Adaptive traffic-signal control using Multi-Agent Deep Reinforcement Learning (MAPPO). Intersections as cooperating agents trained on SUMO simulations of real Bucharest road networks. FastAPI + React + MapLibre GL frontend.

Honeywell-sponsored assistive-navigation prototype for visually impaired users — Raspberry Pi, IMU, BLE beacons, ultrasonic sensors, bone-conduction audio, directional haptics. React PWA, OSRM routing, OpenAI, ElevenLabs, Kalman filtering.

Functional hybrid-daylighting hardware prototype — ESP32, dual-axis solar tracker, Fresnel optics, PMMA optical fibre, complementary LEDs, closed-loop embedded C/C++ control, IoT monitoring dashboard.

High-density Proxmox virtualisation platform — NUMA-aware scheduling, CPU pinning, memory ballooning, huge pages, Docker/Portainer, Tailscale mesh, Cloudflare Tunnel, automated backups and monitoring.

Urban pedestrian-routing platform prioritising shaded, thermally comfortable routes — React Native, Mapbox GL, PostGIS, Redis, OSRM, solar-shadow simulation, Gradient-Boosted Decision Trees for segment comfort scoring.

Computer-vision pipeline for detecting churches in Romanian satellite imagery — U-Net semantic segmentation, ResNet18 transfer learning, YOLOv8n object detection. PyTorch, CUDA, stratified K-fold validation.