AI Research Lab
What we offer
At CertEq, we are building a research lab where the brightest minds in ML can freely pursue their own directions while testing ideas against the hardest real-world system: financial markets. We are a young, independent team without bureaucracy, driven by a mission to rethink how complex systems can be modeled and predicted — starting with markets and extending to global supply chains, energy, and beyond.
About us
For seasoned AI researchers
Foundation models building
Apply if you're excellent in one of following topics:
Foundation models: pre-training, post-training (finetuning, RL), reasoning, inference-time scaling
Time-series architectures: transformers, SSMs, diffusion models, alternative (e.g. reservoir networks)
Multimodal learning: language, image, time-series, tables, graphs and other modalities
Neural scaling laws and emergent behaviors at both training and inference stages
You have strong publication record (NeurIPS, ICML, ICLR).
Apply
For seasoned AI researchers
Foundation models scaling
Apply if you meet the minimum qualifications:
MS/PhD and hands-on multi-GPU pre-training of large foundation models (transformers, SSMs)
Deep experience with FSDP/ZeRO, tensor & pipeline parallelism, mixed precision, gradient checkpointing.
Strong GPU systems fundamentals: NCCL, CUDA/Triton basics, and profilers (nsys/nvprof/torch profiler).
Track record taking models prototype → scaled training → production inference
Apply
For PhD students and Postdoc
Internship program
Publishable research
USD 7,000/month
Direct mentorship from CSO
Option for a full-time offer
Requirements
Current PhD student or postdoc in ML/DL/AI
20 hrs/week and minimum 3 months commitment
Research mindset and ability to work independently
Apply
Irina Rish is a Full Professor at the Université de Montréal (UdeM), where she leads the Autonomous AI Lab, and is a core faculty member of MILA - Quebec AI Institute. She holds a Canada Excellence Research Chair (CERC) and a CIFAR Chair. Irina completed her MSc and PhD in AI at the University of California, Irvine, and holds an MSc in Applied Mathematics from Moscow Gubkin Institute.
About Irina
Core foundation
Multimodal foundation model
Neural Scaling Laws
Efficient Architectures & Inference
HPC & Large-Scale Training
Adaptation & Strategy
Continual & Online Learning
Sequential Decision Making & RL
Synthetic Data & Simulation
Risk Management & Robustness
Automation & Frontier
Interpretability & Mechanistic Insights
Fully automated trading pipelines
Explore beyond transformers/SSMs
Low-level code agent
In addition to conventional hardware, we have access to novel chip architectures. Our research focuses on accelerating low-level software development for these new architectures using AI-driven code generation.