Train models and
read papers in one place.
Five purpose-built AI modes. Thousands of workflows. Fine-tune models on GPU, solve hard math, run evaluations, and publish research — all in one continuous workspace.
Switch instantly. Keep everything.
Research, Biology, Flywheel, Write, and Mathematics — five expert modes in one workspace. Switch mid-session without losing any context.
Research
ModeFull-cycle research automation: from framing hypotheses and synthesizing literature to planning experiments, GPU training, running evaluations, and drafting manuscripts.
Full-cycle research automation: from framing hypotheses and synthesizing literature to planning experiments, GPU training, running evaluations, and drafting manuscripts.
Your specialist for both wet-lab and computational biology — covering protein design, genomics, pathway analysis, and biomedical reasoning workflows.
Turn production usage into compounding model gains — automatically design fine-tunes, run evaluations, and continuously ship better versions.
Transform rough notes into publication-ready papers with structured arguments, fact-checked citations, clean LaTeX, and camera-ready layouts.
Solve anything from arithmetic to research-level mathematics — step-by-step working, KaTeX-rendered proofs, formal theorem proving, function graphs, and web-grounded solutions from MathOverflow and arXiv.
1000s of pipelines.
Ready-to-run pipelines spanning research, biology, model training, and writing — use them as-is or adapt them.
Your Keys.
Your Compute.
Set up once in the dashboard. Your GitHub, Hugging Face, Weights & Biases, and other API keys automatically flow to every agent and session.
Jump straight into research.
Package dependencies, Python environments — all of it is handled by our sandboxed runtime. No setup. No config. Just research.
Agents that keep going.
Persistent runtimes keep long-running workflows alive in the browser. Kick off a training run or literature review, step away, and your agents carry on without you.
Stateful Sandboxes
Every agent runs inside an isolated, stateful environment with full checkpointing and reliable resume.
Flexible Compute Scaling
Go from a single analysis node to a full GPU cluster — no need to change prompts, workflows, or context.
Background Job Orchestration
Send long evaluations, training runs, and literature pipelines to a background queue while you're away.
I created Veil because I kept jumping between five separate tools just to complete a single research task — reading papers in one tab, running models in another, writing in a third. Every switch shattered my flow and wiped my context.
Researchers and ML engineers deserve a workspace where the AI holds the full context of their work — not just the last message. That's what Veil is. One unified workspace, four specialized modes, thousands of workflows, your own compute — all together, always in sync.
Research smarter.
Build stronger models.
Join researchers who are running experiments, reviewing literature,
and training models — all inside one seamless workspace.