计算与实验科学家
Computational and Experimental Scientist
职位描述
这个职位负责早期阶段AI驱动的蛋白质和肽工程公司的完整设计-制造-测试-模型循环：你将运行并改进基于口袋条件的离散扩散模型用于序列设计，并亲自执行闭合循环的结合动力学。你将加入一个精简的核心团队，直接向CEO汇报，这是公司中最具影响力的科学职位之一。
你将负责的工作
- 通过架构优化、新注意力头和分层推理来改进和扩展专有的扩散模型和配套折叠模型。
- 在大规模上运行机器学习推理平台，并诊断注册、流失和客户群体中的使用模式。
- 操作流体处理机器人和板自动化（Hamilton、Tecan、Opentrons或同等设备），并交付可靠、可生产使用的协议。
- 全流程负责BLI和SPR：实验设计、固定化、再生、参考、稀释系列、动力学拟合、质量控制和故障模式诊断。
- 编写详细的云实验室协议并管理内部筛选仪器。
- 从受体生物学和蛋白质结构到评分函数，再到科学家实际使用的平台输出，负责整个技术栈。
- 闭合设计循环：从平台上获取序列，运行动力学，更新模型，并交付改进的序列。
我们寻找的人选
- 2年以上构建或运营离散扩散模型、蛋白质语言模型（如ESM或ProtT5）或结构预测系统，在真实的制造-测试-模型循环中，而非学术论文或公开微调。
- 亲自编写并调试液体处理协议并在机器人平台上部署到生产环境，而非监督核心设施。
- 亲自完成BLI或SPR动力学曲线的全流程拟合并诊断故障模式：质量传输、探针亲和力、非特异性结合、聚集、钩状效应、错误参考。
- 熟练使用Python或等效脚本语言编写机器人方法和分析动力学数据。
- 有闭合完整循环的实际能力：设计序列、合成或表达、测量动力学、更新模型、迭代优化。
- 能够将蛋白质语言模型和序列设计工具（如RFdiffusion或BindCraft同类工具）作为输入和输出，而非黑盒。
- 生物学、生物化学或相关生命科学领域的扎实背景。
- 具备运营思维：善于利用资源，注重行动，能够高效执行。
查看英文原文
ABOUT THE ROLE
This role owns the full design-make-test-model loop at an early-stage AI-driven protein and peptide design company, working directly with the founding team. You will advance pocket-conditioned discrete diffusion models for sequence design, operate an inference platform at scale, and close the loop with hands-on kinetics, making you one of the most end-to-end scientists on a lean core team of five to seven people.
WHAT YOU'LL DO
- Improve and extend discrete diffusion models and companion folding models with refinements, new attention heads, and hierarchical reasoning.
- Operate an ML inference stack at scale and diagnose usage patterns across customer segments, signups, and churn.
- Own fluid-handling robotics and plate automation (Hamilton, Tecan, Opentrons, or equivalent) and ship reliable, production-ready protocols.
- Run BLI and SPR end-to-end: assay design, immobilization, regeneration, referencing, dilution series, kinetic fitting, and QC.
- Write precise protocols for cloud labs and manage internal screening instrumentation.
- Work across receptor biology, protein structure, scoring functions, and sequence design outputs.
- Close the loop: take sequences from the platform, generate kinetics data, update the model, and iterate on improved sequences.
WHAT WE'RE LOOKING FOR
- 2+ years personally building or operating discrete diffusion models, protein language models (e.g. ESM, ProtT5), or structure prediction systems in a real design-make-test cycle.
- Hands-on experience writing and debugging liquid-handler protocols on robotic platforms and shipping them to production.
- 3+ years in a GTM, solutions engineering, or customer success role in biotech or life sciences SaaS, with a track record converting free-tier users to paid tiers.
- Direct, personal wet-lab experience; not limited to supervising core facilities.
- Personally fitted BLI or SPR kinetic curves end-to-end and diagnosed failure modes such as mass transport, tip avidity, aggregation, and hook effect.
- Proficiency in Python for scripting robot methods, automation, and kinetic curve fitting.
- Comfort treating protein language models and sequence design tools (e.g. RFdiffusion, BindCraft) as inputs and outputs, not black boxes.
- Understanding of receptor biology, protein structure, and scoring functions sufficient to diagnose why a predicted ddG failed on a sensor.
- Operator mentality: bias toward direct execution, rapid iteration, and shipping results.
- Background in gene editing, gene therapy, or receptor trafficking is a plus.
- Experience at biotech startups, accelerators, or prior exits is strongly valued.
COMPENSATION & BENEFITS
Initial consulting engagement: $3,000 to $5,000 per month. Full-time conversion: base salary of $80,000 to $200,000 depending on profile, with heavy equity and deal-contingent upside. No visa sponsorship available.
LOCATION
Hybrid, based in New York, NY. Increased on-site presence expected once internal screening instrumentation is operational, anticipated within three to six months.