Atna.AI is building cutting-edge synthetic media detection technology — and we're looking for a self-driven AI Engineer to lead the research.
You won't just call APIs. You'll open up state-of-the-art models, rebuild their internals, and ship what you build into production.
If you want to work at the edge of AI research and actually see it deployed, not just published — let's talk.
Build and fine-tune foundation models for deepfake detection, image-edit segmentation, and video/audio forensic analysis
Work under the hood — Transformer encoders/decoders, attention blocks, latent spaces — not just pre-built pipelines
Push CLIP, Vision Transformers (ViT), and U-Net architectures for cross-modal analysis and precision segmentation
Design custom loss functions (BCE + Dice, contrastive loss) to hit pixel-level detection accuracy
Turn research code into production-grade, optimized Python pipelines
Ship it — containerize with Docker, deploy via GitHub Actions across Swarm clusters
2–3 years of hands-on experience training/fine-tuning foundation models
Strong in Python, PyTorch, NumPy, CUDA
Deep knowledge of ViT, CLIP, Auto-Encoders, U-Net, self-supervised learning
Comfortable with Docker, Docker Swarm, CI/CD, FastAPI/Flask
You can read a dense research paper and turn it into working code — on your own
Bonus: Experience with latent space manipulation, PCA, transfer learning, or blending graph algorithms with neural probability maps