Applied ML Engineer ($170k-$280k + Equity) at Macroscope
Company: Jack & Jill
Location: San Francisco, CA
Type: Full-time
Posted: 2026-09-16
About this role
This is a job that Jill, our AI Recruiter, is recruiting for on behalf of one of our customers.
She will pick the best candidates from Jack's network.
The next step is to speak to Jack.
Job Title
Applied ML Engineer
Salary
$170k-$280k + Equity
Company Description
Macroscope is a San Francisco-based startup building infrastructure for agent-powered software development, backed by Lightspeed, Thrive Capital, and Google Ventures. Founded by repeat exited entrepreneurs from Periscope and Magic Pony, the team is creating tools that allow engineers to direct fleets of agents with automated code review and verification.
Job Description
As an Applied ML Engineer at Macroscope, you will own the machine learning lifecycle for our agentic software development infrastructure. You will focus on training and fine-tuning LLMs using advanced reinforcement learning techniques like GRPO and DPO. By building high-quality evaluation datasets and rigorous benchmarks, you'll drive model improvements that empower engineers with unprecedented leverage.
Location
San Francisco, USA
Why this role is remarkable
- You will perform deep technical work in reinforcement learning (RLHF, DPO, GRPO) and model training, moving beyond simple API wrappers to build core agentic infrastructure.
- Macroscope is led by veteran founders who previously exited Periscope and Magic Pony, and is supported by top-tier investors including Lightspeed, Thrive Capital, and Google Ventures.
- You'll receive significant early-stage equity (0.25%-0.5%) at a valuation that offers massive upside compared to later-stage competitors, working in a high-agency, flat-structured environment.
What You Will Do
- Train and fine-tune LLMs using reinforcement learning techniques such as GRPO, DPO, and PPO to improve reasoning and code generation capabilities.
- Design and curate high-quality evaluation datasets and benchmarks to rigorously measure model perf...