/roles — ROLE_310
MTS - Research Scientist
Small, well-funded AI company training bespoke forecasting models for enterprise supply chains
The role
- COMP
- $200K - $300K
- EQUITY
- Competitive equity
- LOCATION
- San Francisco
- WORKPLACE
- On-site
- EXPERIENCE
- 4+ years
- VISA
- None, Visa transfers
- STACK
- Python, PyTorch, TensorFlow
- INDUSTRY
- AI, Logistics, Manufacturing, Software Development, Enterprise, B2B
The company
Well-funded AI company building forecasting models tailored to every client's supply chain, helping manufacturers and brands decide what to make, store and ship with far less waste.
- STAGE
- growth-stage
- FUNDING
- $60M+ raised
- TEAM
- ~15 people
- FOUNDED
- 2024
- BACKING
- backed by Lightspeed, Kleiner Perkins
JD — the work
About the role
The research team, just a handful of people, owns the core model: a general architecture that blends quantitative forecasting with LLM-based reasoning and is then fine-tuned on each customer's own data. Research capacity directly limits how quickly new customers can be onboarded, so your work matters immediately. You would drive research from idea to deployed model yourself and see it reach production within weeks, with a clear view of how it changes real-world operations.
What you'll do
- Carry each idea from hypothesis to a deployed model and see the real-world operational effect it has
- Build optimization and forecasting models for many-variable problems whose goals move as business conditions change
- Curate in-house datasets that capture meaningful signal from actual supply chains and commercial activity
- Fold LLM reasoning into specialized numerical models so forecasts become more accurate and adaptable
- Keep models current as launches, shifts in consumer mood and geopolitical events change demand
What they're looking for
- 4+ years of hands-on research experience in machine learning or a related quantitative field
- Background in forecasting, optimization, deep learning or reinforcement learning
- Fluency in Python with PyTorch or TensorFlow
- Appetite for owning research from start to finish and seeing it in production quickly
- Interest in combining language models with purpose-built quantitative methods