About us:

DevKit is more than just an IT company - we are a team of passionate professionals dedicated to creating cutting-edge solutions. Our employees work on exciting international startups, collaborating with highly skilled teams from around the world. With expertise in software development, design, and engineering, we help bring innovative ideas to life and drive success on a global scale! Our mission is to connect exciting projects with top-tier professionals. We are committed to creating a dynamic space where innovation thrives, top talent excels, and global teams unite to develop cutting-edge solutions.

Project:

We are building specialized AI/LLM models for software security and code analysis. The models will analyze real-world vulnerabilities, perform security audits, identify exploitable code, and generate secure remediation and refactoring recommendations.

About the role:

We are looking for a Senior Deep Learning / LLM Engineer to contribute to the development and improvement of internal LLMs focused on cybersecurity, security auditing, vulnerability analysis, and secure code analysis. The key requirement is a deep technical understanding of modern LLMs and systems such as Claude. We are not looking for someone who simply uses Claude Code, Codex, or similar tools — the candidate should understand how these models work, are trained, evaluated, and improved.

Responsibilities:

• Develop, train, fine-tune, and evaluate LLMs.
• Build and curate high-quality datasets for LLM training.
• Work with Transformers, Deep Learning, SFT, RLHF/RLAIF, embeddings, and model evaluation.
• Analyze model behavior, failures, reasoning, and code-generation capabilities.
• Improve LLM performance for cybersecurity and code-analysis tasks.
• Build training and evaluation pipelines.
• Work with security experts to transform cybersecurity knowledge into training data.

Requirements:

• Strong Deep Learning / Machine Learning background.
• Hands-on experience with LLMs and Transformer architectures.
• Deep understanding of how models such as Claude operate.
• Experience with LLM training, fine-tuning, evaluation, and dataset preparation.
• Strong Python and modern ML/DL frameworks.
• Understanding of tokenization, attention, embeddings, inference, context management, and post-training techniques.
• Experience with agentic/tool-use or coding LLMs is highly valuable.

Working terms:

A long-term role in a stable, fast-growing company. Flexible working hours with a results-oriented approach. Competitive salary and benefits package (vacation, sick leave, national holidays). Comfortable workspace, corporate events, and team-building. Continuous learning, professional growth, and skill development.

Key profile:

Right candidate should have a strong understanding of LLM architecture, training/fine-tuning, dataset creation, and model evaluation.