We’re an early-stage AI company building a new infrastructure layer for AI systems: enabling models and agents to reliably interact with the real-world data that matters.
Today, most enterprise data (~80%+) is unstructured and multimodal—documents, tables, charts, images—yet existing systems (LLMs, OCR, RAG pipelines) fail to process it reliably.
We’re solving this by building state-of-the-art vision-language models and systems that transform complex, real-world inputs into structured, machine-readable representations that downstream models can actually reason over.
Our models are already deployed in production across high-stakes domains (finance, legal, healthcare) where accuracy matters, and are outperforming existing approaches on real-world benchmarks.
We believe this layer—making messy reality legible to AI systems—is one of the most important unsolved problems in the stack.
The Role
We’re hiring a Founding ML Researcher to work directly with the founders to define and build the core research agenda.
This is not a typical applied ML role. You’ll be operating at the intersection of:
- Multimodal foundation models
- Vision-language reasoning
- Structured generation / parsing
- Reliability and determinism in AI systems
You’ll have end-to-end ownership—from first-principles research → model design → training → production deployment.
What You’ll Work On
- Designing novel architectures for multimodal understanding (documents, tables, layouts, graphs)
- Pushing beyond standard LLM paradigms into structure-aware and layout-aware models
- Improving factuality, determinism, and reliability in model outputs
- Building systems that combine:
- Vision models
- Language models
- Structured decoding / constrained generation
- Developing evaluation frameworks for real-world correctness (not just benchmark scores)
- Shipping research directly into production systems used by real customers
Who This Is For
We’re specifically looking for researchers currently at (or competitive with):
- Frontier labs (e.g. Anthropic, OpenAI, DeepMind, Meta, etc.)
- Top-tier research groups or high-end startups doing foundational ML work
You should have:
- Strong background in deep learning / ML research
- Experience with at least one of:
- Multimodal models (VLMs, vision transformers, etc.)
- LLMs / generative models
- Representation learning or structured prediction
- A track record of building or shipping real systems, not just papers
- Taste for first-principles thinking over incremental work
What Makes This Different
- Greenfield research direction: You will define major parts of the roadmap
- Tight feedback loop: Your work goes into production quickly
- Hard, unsolved problems:
- Turning perception into structured reasoning
- Bridging vision + language + symbolic structure
- Making AI systems reliable in the real world
- Small, elite team: You’ll work directly with highly technical founders
- Massive surface area: This problem sits upstream of RAG, agents, and enterprise AI
Why Join
Most frontier labs are focused on scaling general models.
We’re focused on something orthogonal and equally critical:
Making models actually work on real-world data.
If solved, this unlocks:
- Reliable AI agents
- Production-grade automation
- Entire categories of vertical AI
This is a chance to own a foundational piece of the AI stack from day one.
