🏠 My Workbench

⚗️ My Experiments

Every experiment result you submit (real yields from runs or literature) is stored here under your account. Validated experiments can be "reapplied" — injecting their points as training data into the engine's Gaussian process to generate the next round of optimization suggestions, and earn credits.

🔁 Reapply

📊 Data Import + Auto-Fit

Upload experiment data (CSV / Excel export) and auto-fit Monod / Logistic / Andrews / Haldane growth-kinetic models, returning parameters + R² + error bars. No coding required.

📋 Data preview

🏆 Fit results (ranked by R²)

R² > 0.95 = excellent fit · R² > 0.85 = usable · R² < 0.85 = add points or switch model. Parameters include error bars (95% CI).

🪙 Contribute Credits

Other platforms only deduct credits; here, you earn credits by contributing real experiment data. Earned credits offset compute cost, forming a "data → credits → compute" flywheel.

📈 Credit earning rules

• Submit a feedback with a real result (success or failure): +10 credits
• Feedback judged "validated" (actual close to predicted, or signal success/improved): +30 credits
• First experiment submission on an engine: +20 credits extra
• Submit a negative sample / failure case: +15 credits (a scarce asset, rare in the industry)
• Your validated data reapplied / cited by others: milestone bonus rewards

🏆 Contributor leaderboard

Ranked by cumulative earned credits (emails anonymized). Want to be listed? Run experiments in the Lab and backfill real results.

🔑 My API (bring your own LLM)

Hand your own model key to the platform: routed through your key, zero platform cost, your data never leaves your account. Ideal if you already have DeepSeek / Qwen (DashScope) / OpenAI / local Ollama quotas.

🧪 Try your own API

The request below is forwarded via POST /api/v2/llm using your key; the platform spends no credits. Save your key first.

💡 Getting the most value

• After generating a plan in the Lab's "virtual experiment", use your key for natural-language polishing / interpretation, with zero platform credit cost.
• Data privacy: prompts and results are sent only to your configured provider; the platform keeps nothing.
• Change the key / model anytime via the fields above; removing it restores the platform default (if configured).