AI Engineer

Spaarkly s.r.l.; Town of Italy (NY); On-site; USD 100,000 - 130,000; Full time

Benefits offered by this job

Permanent employment contract (CCNL Commercio)

Remote opportunity

Company welfare policy

Professional development support

Access to cutting-edge tools

Job summary

Spaarkly s.r.l. is hiring for an AI engineering position focused on developing and fine-tuning generative 3D models and enhancing segmentation pipelines for their Virtual Try-On system. Candidates should be proficient in Python, generative AI, and familiar with deep learning frameworks. The role offers remote work opportunities with periodic onsite alignments in Potenza, benefits including health and wellness support, and professional development initiatives.

Qualifications

• 2+ years of relevant experience or strong portfolio of personal/open source projects.
• Hands-on experience with fine-tuning (LoRA, DreamBooth, etc.).
• Ability to design, clean, and manage training datasets.

Responsibilities

• Fine-tune generative 3D models for eyewear dataset.
• Develop segmentation and matting pipelines for Virtual Try-On.
• Take ownership of AI integration across workflows.

Skills

Python

Generative AI & Diffusion Models

Stable Diffusion

Dataset curation

Deep learning frameworks (PyTorch, TensorFlow)

Git

Tools

TripoSR

Gaussian Splatting

Computer vision techniques (SAM, YOLO, rembg)

Job description

We’re opening an AI engineering position. You’ll work directly on the models and pipelines that power our core products — with real ownership, real impact, and a clear roadmap ahead.

What you’ll work on

Two concrete projects define the 2026 roadmap:

• Fine-tuning generative 3D models on our existing eyewear dataset to automate the creation of photorealistic 3D assets from product images
• Developing and optimizing segmentation and matting pipelines for our live Virtual Try-On system

Beyond these, you’ll progressively take ownership of AI integration across our workflows: RAG pipelines, MCP server design, automation of internal processes, and LoRA-based post‑training on new domains.

Required Skills

• Python — solid, production‑oriented proficiency
• Generative AI & Diffusion Models — hands‑on experience with fine‑tuning (LoRA, DreamBooth, or equivalent techniques)
• Stable Diffusion, Hugging Face ecosystem, or similar
• Dataset curation — ability to design, clean, and manage training datasets
• Deep learning frameworks — PyTorch and/or TensorFlow
• Git — standard version control in collaborative environments
• 2+ years of relevant experience, or a strong portfolio of personal/open source projects that demonstrates equivalent depth

Bonus skills

• 3D generative models — experience with TripoSR, Zero123++, InstantMesh, Gaussian Splatting, or NeRF‑based approaches
• ComfyUI — workflow design and custom node development
• Real‑time inference optimization — model quantization, ONNX, TensorRT, or edge deployment
• RAG pipelines — LangChain, LlamaIndex, or equivalent
• MLOps — experiment tracking (MLflow, W&B), cloud training (GCP, AWS, RunPod)
• Computer vision — segmentation, matting, object detection (SAM, YOLO, rembg)

Ideal candidate

• You have a strong core in generative AI and fine‑tuning — and the intellectual curiosity to grow into adjacent areas (3D generation, real‑time inference, RAG, automation)
• You are comfortable working with imperfect data and open research problems
• You work well autonomously, but know when to align with the team
• You follow the open‑source AI community — Hugging Face, ComfyUI, research papers — not just as a user, but as a practitioner
• You want to build something vertical and proprietary, not just integrate third‑party APIs

Benefits

• Permanent employment contract (CCNL Commercio)
• Remote opportunity — we prefer candidates based in Southern Italy or neighbouring regions, with monthly on‑site alignment in Potenza
• Company welfare policy including health and wellness benefits
• Work devices and tools provided
• Professional development support: training, certifications, and industry conferences
• Access to cutting‑edge tools and compute infrastructure

Working mode

In‑office | Hybrid