How We Found an Architect of Autonomous Research Platform for 42.
Valeriya
Lead IT recruiter in Lucky Hunter
Veronika
Sourcing IT Specialist in Lucky Hunter
About the company
42. is a young research startup developing fundamental ML models for analysing financial markets. The company benefits from the scientific leadership of a Professor from the Université de Montréal and Mila – Quebec AI Institute, with its engineering and research teams driving ambitious R&D projects.
< 20
Employees
Canada
Headquarters
Building ML models for financial analysis
Key Feature
This is one of several roles we've filled for 42. As a deep-tech agency, Lucky Hunter handles high-complexity roles where standard filters and basic sourcing simply don't work.
Rare Talent: How We Secured 3 Low-Level Engineers for 42.
The Entry Point
Rather than hiring a team of quants, 42 chose to automate their HFT pipeline end-to-end with AI agents: from trade idea generation to validation, deployment, and optimization. That meant finding an architect who could build the system from scratch, not a standard ML Researcher.
Target Profile
- LLM architectures: RAG, fine-tuning, prompt engineering, evaluation frameworks
- Multi-agent systems: Orchestration, memory, tool use, agent collaboration
- Hands-on experience: A track record of building auto-researcher / co-scientist systems – autonomous research agents assisting scientists or analysts
- Core stack: Strong Python and ML stack (PyTorch, JAX)
- Experimental design & stats: Experimental design and statistics for non-stationary, noisy environments
- Systems thinking: Abstractions, interfaces, pipelines – not just models
- Languages: English B2+ level
- Nice to have: HFT/MFT experience, low-latency systems
Challenges
The main difficulty lay in the overlap of requirements. We needed someone who had actually built autonomous research agents or AI co-scientist platforms before. Globally, there are only a handful of people like that on the market.
On top of that, there was no playbook. These kinds of systems are brand new, so there were no existing frameworks to follow. Trading experience was a major plus (though not a hard requirement), it gave candidates immediate context on the operating environment.
On top of that, there was no playbook. These kinds of systems are brand new, so there were no existing frameworks to follow. Trading experience was a major plus (though not a hard requirement), it gave candidates immediate context on the operating environment.
The Lucky Hunter Method
We relied heavily on the candidate pipeline we had already built: by the time this role opened, we had an active LLM/AI-agent pipeline for 42. with several warm candidates from related searches.
We re-evaluated those profiles through the lens of the new role and found an exact match – a candidate from a different search had already built a co-scientist-style system for processing hypotheses and had trading experience on top of it, a rare combination for this market.
We re-evaluated those profiles through the lens of the new role and found an exact match – a candidate from a different search had already built a co-scientist-style system for processing hypotheses and had trading experience on top of it, a rare combination for this market.
1,5 month
Time to Close
105
Resumes Screened
3
Vacancies Closed
The candidate went through just two interview rounds with the CPO and Founder before getting an offer. While the role originally required relocation to Amsterdam or Dubai, 42 agreed to a remote setup from Spain to secure the candidate.
ML Researchers with Niche Expertise: Where We Searched and How We Found 3 Experts for 42.
Results
We already had great momentum with our AI agent candidates. As soon as I looked at his background, the fit was obvious: at his previous role, he was building something very close to an AI co-scientist, and he had a solid background in trading. 42. team moved fast and stayed flexible – approving remote work from Spain to secure a candidate this rare.
Recruiter’s Feedback
Valeriya
Lead IT recruiter in Lucky Hunter
Share
Recent Cases