TixelJobs
E
EverAIvia muse

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

REMOTEPosted 1w ago
ML EngineerSeniorFull-time#remote

About the Role

**Our Vision \& Products**
==========================

***EverAI — Building the Future of AI Companionship***

*One of the Top 15 Largest \& Fastest\-Growing AI Companies in the World*

*50 Million Users in 2 years — Help Us Reach 100M first, 500M next*


At **EverAI**, we’re shaping what it means to connect with AI. With **50 million users and counting**, we're not just building products — we're creating entirely new categories.


Our flagship product is the **world’s largest AI companionship platform**, redefining relationships for millions. It is governed by our proprietary moderation system, EverGuard — an internal AI designed to ensure everything we build is **safe, ethical, and human\-first**.

*And we’re only just getting started!*

**Our Team**
============


We are an enthusiastic, passionate and hardworking team of 100 people. Our founding team has strong entrepreneurial experience building and scaling web products from 0 to IPO.

**Alexis Soulopoulos** \[CEO]

* *10\+ years in Tech Executive Leadership*
* *Co\-Founder Mad Paws Holdings (from 0 to IPO)*
* *Forbes 30 under 30 \+ Deloitte TechFast50 ’22 \& ‘23*

**Michael Monin** \[Co\-founder \& CTO]

* *10\+ years as CTO / COO (web2/web3\), 3 years in AI/LLM*
* *Serial\-entrepreneur: MTK Digital (exited / 0\-\>$20m revenue) and Zipchat (AI Chatbot for E\-commerce brands)*

**Thomas Lacroix** \[Co\-founder \& CMO]

* *8\+ years in Customer Acquisition \& E\-commerce Growth*
* *Serial\-entrepreneur: Curatible (sold to Blackstone) and MTK Digital (exited / 0\-\>$20m revenue)*

**Maruša Fasano** \[CFO/Legal]

* *25\+ years in Finance, Strategy, M\&A*
* *Ex\-CFO/M\&A @Curatible (exited to Blackstone)*
* *Ex\-President of the Board @SotremoSA (exited)*
* *Co\-founder/CFO @SoftOne (exited)*

**Your Role**
-------------


Most LLM roles stop at experimentation.

**Here, the models you improve power experiences for 50M\+ users.**


Most of our traffic runs on **self\-hosted, fine\-tuned open\-weight models**. On our internal roleplay evals, our smaller fine\-tuned models can outperform much larger off\-the\-shelf models. **Your job is to make them better.**


You’ll join a small AI team led by **Marco, our LLM Tech Lead**, working closely with our CTO, Product, Web and DevOps teams. You’ll work across **data, post\-training, evals, and model behavior,** taking ideas from failure analysis to production.


The problems are real: **personalization, character consistency, long conversations, better evals, multilingual behavior, and faster training cycles.**

**If you want to shape model behavior at consumer scale, this is the role.**

**Key Responsibilities**
------------------------

* **Fine\-tune open\-weight LLMs** using SFT, LoRA/QLoRA and post\-training methods such as DPO and GRPO
* **Diagnose model failures and turn them into experiments.** Decide whether the answer is better data, post\-training, prompting or classifiers
* **Build training data and evals** that directly shape what gets trained next
* **Improve model behavior** across personalization, consistency, tone, multilingual and group interactions
* **Ship to production** and use real user behavior to guide the next iteration
* **Speed up the training loop** so experiments are faster, easier to compare, and more reproducible

**Your Qualifications**
=======================

* You’ve **personally post\-trained open\-weight LLMs** for a user\-facing product and shipped them to production
* **You’ve worked where AI was core to the product or business**, not an experimental side feature
* **Fine\-tuning** is core to your experience
* Strong **Python, PyTorch and Hugging Face**, with hands\-on GPU training and debugging
* You **build data and evals yourself** and make decisions from evidence
* You can take work **beyond notebooks into production**
* You **use AI tools daily** in your own engineering work
* **You are fully comfortable working with uncensored models and explicit NSFW content**, including harmful materia
Share