About Charlemagne Labs
Charlemagne Labs is building the first Human Digital Hygiene platform—an intelligent, on-device safety agent ("Agent Charley") that prevents humans from making costly cybersecurity mistakes. Our mission is to defend individuals and organizations against the new wave of agentic, AI-driven social engineering threats. We're combining cutting-edge small language models (SLMs), privacy-first system design, and HCI research to create proactive, zero-egress defenses that stop bad clicks before they happen.
Role Overview
We're looking for an AI Engineer who's excited to work at the intersection of applied machine learning and cybersecurity. You'll help fine-tune and evaluate local LLMs that detect, reason about, and intervene in risky human-computer interactions (e.g., phishing, spoofed sign-in pages, or malicious URLs). You don't need a background in cybersecurity—but curiosity about the threat landscape, adversarial behavior, or human error is a big plus. We are looking to hire immediately and move quickly for the right candidate.
Responsibilities
- Fine-tune open-source language models (Gemma, LLaMA, Mistral, Qwen, etc.) for domain-specific reasoning about browser-based risk
- Build and maintain pipelines for dataset preparation, tokenization, supervised fine-tuning (SFT), and evaluation
- Experiment with parameter-efficient techniques (LoRA, QLoRA, or adapters) for small-device deployment
- Contribute to labeling, prompt engineering, and evaluation of threat detection capabilities (e.g., phishing intent or anomaly detection)
- Develop safety classifiers or scoring systems that map model reasoning to user-facing confidence levels or 'amber alert' thresholds
- Collaborate on experiments studying local inference, hallucination reduction, and privacy-preserving architectures (e.g., zero-egress LLM operation)
- Integrate the SLM into browser extensions and local agents
- Build and maintain training datasets reflecting realistic user interactions and cyberattack patterns
- Implement and monitor model performance metrics and contribute to model interpretability research
Preferred Qualifications
- Bachelor's degree in Computer Science, Data Science, or related field, or equivalent experience
- Hands-on experience with Python, PyTorch, and Hugging Face Transformers
- Experience with model fine-tuning or instruction-tuning workflows
- Familiarity with LLM evaluation metrics and prompt engineering
- Interest in cybersecurity domains such as phishing, anomaly detection, or malware classification (optional)
- Interest in AI safety, privacy-preserving ML, or human-computer interaction for cybersecurity
What You'll Learn
How small, local LLMs can power real-time cybersecurity defense. How to apply AI reasoning to detect and interrupt advanced social engineering attacks. How agentic AI systems are reshaping both offensive and defensive cyber operations. Startup-level productization of research: moving from proof-of-concept to deployed, safety-critical software.
Why Join Us
You'll be joining a small, mission-driven team of builders from Meta, DoD, and cybersecurity research who are reimagining digital safety for the era of agentic AI. You'll work directly with our founder and lead engineer, contribute to published research, and help define a new category: Human Digital Hygiene.
