Mid/Senior AI LLM Engineer

This is a fully remote position, open to applicants in Poland, +1 more country.

📋 Description

• Refine open-weight large language models (LLMs) utilizing SFT, LoRA/QLoRA, DPO, and GRPO methodologies.

• Identify model failures and convert them into experimental opportunities.

• Assess whether enhancements necessitate improved data, post-training adjustments, prompting strategies, or classifiers.

• Create training data and evaluations that will influence future training initiatives.

• Enhance model performance in areas such as personalization, consistency, tone, multilingual capabilities, and group interactions.

• Deploy models to production and leverage actual user behavior to inform subsequent iterations.

• Accelerate the training cycle for quicker, more comparable, and reproducible experiments.

• Collaborate with various teams including AI, CTO, Product, Web, and DevOps.

• Contribute to the enhancement of AI companionship experiences utilized by over 50 million users.

⛳️ Requirements

• Experience in post-training open-weight LLMs for user-facing products and successfully deploying them to production.

• Background in environments where AI is integral to the product or business model.

• Expertise in fine-tuning is essential to your professional experience.

• Proficient in Python, PyTorch, and Hugging Face.

• Practical experience with GPU training and debugging.

• Ability to independently build training data and evaluations, making decisions based on evidence.

• Skilled in transitioning work from notebooks to production environments.

• Regularly utilize AI tools in engineering tasks.

• Comfortable handling uncensored models and explicit NSFW content, including challenging materials when developing and assessing moderation systems.

• Experience with conversational AI, roleplay applications, or highly personalized consumer AI is advantageous.

• Familiarity with multilingual LLMs is a plus.

• Knowledge of moderation, safety protocols, or adversarial testing is beneficial.

• Experience with multimodal models is a plus.

• Contributions to open-source LLMs or post-training initiatives are advantageous.

🏝️ Benefits

• 4 weeks (20 working days) of paid vacation annually.

• Fully remote work setup.

• Annual in-person gathering.

• Monthly allowance of 100 USD for health insurance costs.

• Unlimited one-on-one sessions with psychologists and lifestyle consultants via OpenUp, available to up to three family members.

• Co-working space budget for up to two visits per month (35 EUR / 40 USD per visit).

• Learning budget allocated for courses, books, conferences, events, or certifications.

• Company-provided laptop.

• Monitor budget of up to 250 USD.

• Budget for keyboard, mouse, microphone, and headphones up to 150 USD.

• Premium access to ChatGPT, Hugging Face, Claude Code, and other essential tools.

• Compensation package includes any applicable target bonuses.