Senior AI Researcher
At a glance
Senior AI Research role at LatamCent. Brazil · full-time · $80,000–$90,000 base.
aimlroles summary, based on the employer's posting.
What you'll do
- Choose promising models, research papers and methods, then test their value with evidence
- Build agents using frontier model APIs to navigate unfamiliar software and handle business tasks
- Create benchmarks and evaluation methods that show whether agents have improved
- Investigate agent breakdowns and use what you find to guide fixes
- Develop prototypes and demos, pass successful approaches to engineering, and share findings through writing and talks
What you bring
- At least five years owning technical results at startups; standout applicants with three or more years may be considered
- Hands-on experience building multi-step agents with frontier LLMs, including tool use and orchestration
- A record of AI products or systems reaching real users, plus disciplined experiment design using baselines and evaluation sets
- Strong spoken and written English, and openness to an independent contractor arrangement paid in USD
Who this fits
This suits a self-directed researcher who can set priorities, move experiments forward, and flag blockers early. The role is hybrid in São Paulo, with three on-site days each week, and follows Brazil time. It is a contract position.
From the employer
The Opportunity
You'll join a small frontier AI team inside a fast growing B2B SaaS company, where the job is figuring out what new models can actually do and turning that into agents customers use. In your first 90 days you'll set your own research agenda and build the benchmarks the team trusts; within a year your experiments will be shaping the product. If you like owning hard, open problems with no backlog and real impact, this is that seat.
What You'll Own
· Decide which new models, papers and techniques are worth betting on, and prove it with data
· Build agents on frontier APIs that understand unfamiliar software and complete real business tasks
· Create the eval and benchmark system that tells the team when an agent truly got better
· Turn agent failures into fixes by digging into where and why things break
· Ship frontier prototypes and demos that leadership shows to customers, then hand the winners to engineering
· Be the team's public voice on what you learn: blog posts, talks and short videos
What We're Looking For
· 5+ years at startups owning technical outcomes (3+ considered for standout profiles)
· Hands on work building multi step agents with frontier LLMs (tool use, orchestration)
· A track record of AI products or systems that reached real users
· Rigorous experiment design: baselines, eval sets, telling signal from noise
· You move work forward on your own and raise blockers early, without needing structure
· AI tools are part of your daily workflow, and you can name what you use for what
· Strong spoken and written English for presentations, travel and writing
· Open to an independent contractor setup paid in USD
Nice to Have
· Time at a top AI lab, or experience training agents in real environments
· RL, post training or fine tuning experience
· Published research, benchmarks, open source work or technical writing
· An active voice in the AI community (posts, talks, GDE or similar)
Tools and Stack
Python, frontier model APIs (Anthropic, OpenAI, Gemini), agent and eval tooling of your choice. The team has no fixed stack and builds its own when that works best.
Details
· Location: Hybrid, São Paulo, Brazil (3 days/week on site); relocation support available
· Time Zone: Brazil (BRT)
· Salary: $6,650 to $7,500/month USD ($80K to $90K/year)
· Equity: Included
· Type: Contract
About the Company
Post Series A B2B SaaS company of about 35 people building an enterprise AI agent that learns any web app on its own and takes action through its UI and API. You'd join its frontier research team, which explores new AI capabilities and feeds them into the core product. Perks include a learning allowance and budget to try the latest AI models and tools.