Mid/Senior AI LLM Engineer (Full Remote - Italy)

In breve

EverAI is a fast‑growing AI company that builds large‑scale AI companionship products for millions of users. Its flagship platform connects people with AI characters and is protected by an internal moderation system called EverGuard. The firm, backed by a leadership team with IPO experience, currently serves about 50 million users and aims to reach 100 million and then 500 million. The opening is for a Mid/Senior AI LLM Engineer who will work remotely from anywhere in Italy, joining a small AI team led by the LLM Tech Lead and collaborating closely with the CTO, product, web and DevOps groups. The engineer will spend each day improving the performance of self‑hosted, fine‑tuned open‑weight language models that power the consumer‑facing product. Core duties include fine‑tuning models with techniques such as SFT, LoRA/QLoRA, DPO and GRPO; diagnosing failures and turning them into experiments; deciding whether improvements require better data, post‑training methods, prompting or classifiers; building training data and evaluation suites that directly influence the next training cycle; and shipping updated models to production where real user behavior informs further iteration. Additional responsibilities involve accelerating the training loop, ensuring experiments are reproducible, and enhancing model behavior in areas like personalization, character consistency, tone, multilingual interaction and group dynamics. Candidates must have hands‑on experience post‑training open‑weight LLMs for a user‑facing product and delivering them to production. They should have worked in environments where AI is central to the business, not merely experimental. Proven expertise in fine‑tuning, strong Python, PyTorch and Hugging Face skills, and daily use of AI tools are required. The role also demands the ability to build data pipelines and evaluation frameworks, move work from notebooks into production, and comfortably handle uncensored and explicit content while developing moderation systems. Preferred, but not mandatory, skills include experience with conversational or role‑play AI, multilingual LLMs, moderation or safety testing, multimodal models, and contributions to open‑source LLM or post‑training projects. The position offers a fully remote setup, four weeks of paid time off, a yearly in‑person gathering, health‑wellness allowances, co‑working space budget, learning budget, and equipment provisions. Premium AI tool access and a culture of ownership, velocity and radical creativity round out the offering. What the role asks for: - Post‑trained open‑weight LLMs for a user‑facing product - Delivered LLMs to production - AI core to product or business - Fine‑tuning as core experience - Strong Python, PyTorch, Hugging Face skills - Built data and evals independently - Comfortable with uncensored/NSFW content - Bonus: conversational AI or role‑play experience - Bonus: multilingual LLM work - Bonus: moderation, safety or adversarial testing - Bonus: multimodal model experience - Bonus: open‑source LLM/post‑training contributions