Protent
AI Research Scientist
United States (on-site or remote) · Full-time · Posted Oct 29, 2026
About the job
Fine-tunes and deploys vision-language models that detect critical activity in live surveillance video and alert operators.
What you'll do
- Fine-tune models and build curated training datasets and evaluation benchmarks from camera footage.
- Improve detection accuracy, latency, and reliability across varied camera conditions.
- Work with engineering teams to deploy model updates and use customer feedback to improve production performance.
What they ask for
- Hands-on experience fine-tuning large models, including VLMs or multimodal models.
- Practical knowledge of LoRA, supervised fine-tuning (SFT), and modern training workflows.
- Experience with video understanding, computer vision, or multimodal AI.
- Python proficiency and familiarity with GPU training infrastructure.
Nice to have
- Published work or shipped projects involving VLMs, video understanding, or fine-tuning at scale.
- Experience with efficient fine-tuning on limited compute budgets or ML in high-stakes or real-time settings.
Tools
LoRA, supervised fine-tuning (SFT), PyTorch, Hugging Face, Python, QLoRA
AI focus
Fine-tuning and evaluating vision-language models for real-time activity detection on surveillance video, then deploying them to live feeds.
Model-written summary · the original posting is authoritative