About the company
Cantina Labs is a social AI company developing advanced real-time models that push the boundaries of expression, personality, and realism. Their technology brings characters to life, transforming how people tell stories, connect, and create.
2023
Founded
San Francisco, USA
Headquarters
Low-latency real-time video models that turn interactive characters into life-like conversational partners.
Key Feature
Main Task
The Goal was to find an Applied ML Engineer experienced in training and scaling large video generative models – end-to-end, from data preparation and training to model distillation, optimisation, and deploying low-latency production solutions.
Who We Were Looking For
- 2+ years of experience building and shipping ML systems with clear ownership.
- Strong PyTorch & Python skills, equally fluent in both training and inference.
- Proven track record of training or scaling generative models, ideally video generation (diffusion / transformers / VAE).
- Hands-on experience with distributed training and large compute runs (DDP / FSDP / DeepSpeed).
- Demonstrated success in performance optimization (latency, memory, cost) and profiling.
- Product mindset: ability to bridge the gap between research concepts and working production code.
Challenges
- Startups, where engineers had relevant experience but worked on smaller GPU infrastructure than required.
- Big Tech, where engineers had the necessary scale, but their compensation expectations exceeded the budget.
The core difficulty was the rare intersection of requirements: the client needed someone with hands-on experience training and optimising state-of-the-art video generation models on massive GPU clusters. Candidates with this niche skillset were scarce and largely split between two extremes:
Our solution
To fine-tune our search, we set up a calibration call with an engineer from the client's technical team. Together, we reviewed several candidate profiles to refine the ideal candidate persona and identify bottlenecks in the recruitment pipeline.
From there, we focused on specific companies known for this kind of expertise, shifting from a broad search to a targeted one.
From there, we focused on specific companies known for this kind of expertise, shifting from a broad search to a targeted one.
2 months
from search kickoff to offer
258
resumes reviewed
1
vacancy filled
Results
ML Researchers with Niche Expertise: Where We Searched and How We Found 3 Experts for 42.
Getting on a call with a tech team member made all the difference. Walking through several candidate profiles together gave us a super clear picture of the ideal fit. Instead of just churning through volume, it allowed us to narrow our focus directly onto key target companies.
Recruiter’s Feedback
Daria
Sourcing IT Specialist at Lucky Hunter
Поделиться
Recent Cases