New opportunity

AI Platform Engineer

Nabla Bio · Boston, United States

About this role

Nabla Bio is hiring a senior AI Platform Engineer to build infrastructure and tools that bring its drug-design AI models into use by AI researchers, wet-lab scientists, collaborators and pharmaceutical partners. The role spans frontend, backend and ML infrastructure, with the goal of shortening design cycles and improving feedback between models and experiments. The posting lists Boston, MA, USA as its location and separately specifies that the role is in person in Cambridge, MA.

Responsibilities

Build and maintain user-facing applications, APIs and dashboards for drug-design workflows.
Develop ML training and inference systems for model development, and serve models through scalable backend services.
Work with scientists to identify problems and deliver tools for AI-augmented experiment planning and analysis. Use LLMs to help propose experiments and learn from results; the company reports measuring 1 million drug designs each month.
Own full-stack delivery from prototype through production, including deployment and observability, and use modern AI development tooling.

Requirements

At least 5 years of experience as a full-stack, ML infrastructure or platform engineer.
Experience building production backend systems that serve ML models; strong frontend development skills, including React and TypeScript; deep fluency in Python; and familiarity with cloud-native tools and containerization, including Kubernetes and Docker.
Strong product judgment, ability to partner with wet-lab scientists and AI researchers, and a record of shipping quality work quickly.

The full-time, in-office role offers access to ML models and wet-lab infrastructure, plus a competitive salary, equity and benefits package. No salary amount is stated.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

PythonReactTypeScriptKubernetesDockerFull-stack developmentFrontend developmentBackend developmentML infrastructureML model servingML training pipelinesML inference pipelinesAPI developmentCloud-native toolsContainerizationDeploymentObservabilityLLMsExperiment planningScientific collaboration

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Make drug-design workflows usable
The posting calls for user-facing applications, APIs and dashboards that shorten the path from an idea to implementation. Prepare to explain how you would turn a scientist’s workflow into a usable tool, including the choices behind its interface and API.
Productionize ML models
The role includes training and inference systems and scalable backend services for serving models. Prepare an example of how you moved a model-backed service into production, focusing on reliability and the boundary between the model and the application.
Connect models with experiments
Nabla Bio wants close collaboration with wet-lab scientists and AI researchers, including tools for AI-augmented experiment planning and analysis. Prepare to describe how you would uncover a scientific pain point and use experimental results to improve the workflow.
Own delivery across the stack
The engineer is expected to take work from prototype to production, including deployment and observability, while shipping quickly with quality. Prepare to discuss decisions across frontend, Python backend and cloud-native deployment rather than treating them as separate handoffs.

Likely challenges to think through

Speed versus production quality
Likely preparation challenge, not a confirmed problem: rapid scientific prototyping may compete with dependable services. Be ready to explain how you decide what to validate early, what to monitor, and when a prototype is ready for production use.
Useful LLM suggestions versus experimental evidence
Likely preparation challenge, not a confirmed problem: an LLM can help propose experiments, but its output needs to be useful to scientists and informed by measured results. Prepare a framework for reviewing suggestions and incorporating feedback without assuming an LLM answer is a validated finding.
One platform, different users
Likely preparation challenge, not a confirmed problem: AI researchers, wet-lab scientists, collaborators and pharma partners may need different views of the same workflow. Prepare to explain how you would prioritize shared infrastructure while adapting tools to concrete user needs.

Practice your answers

Technical practice

How would you design an API that connects drug-design models to a scientist-facing workflow?

What to coverThe role includes building APIs and tools that turn models into usable drug-design workflows. Preparation suggestion: explain how you would define inputs, outputs, validation and errors, and how you would keep the interface usable as models evolve.

How would you build a React and TypeScript dashboard for AI and wet-lab scientists?

What to coverThe posting calls for strong frontend skills and user-facing dashboards. Preparation suggestion: discuss how you would learn what users need to see, represent uncertain or incomplete results, and balance responsiveness with clarity.

How would you develop a Python training pipeline that supports rapid model experimentation?

What to coverThe role includes robust ML training systems, and Python fluency is a qualification. Preparation suggestion: cover reproducible runs, data and model versioning, failure recovery, and the trade-off between iteration speed and reliability.

How would you serve an ML model through a production backend?

What to coverProduction model-serving experience and scalable backend services are explicit requirements. Preparation suggestion: describe request handling, model loading, capacity planning and safe updates, including how you would diagnose slow or failed inference.

How would you deploy and operate these services using Docker and Kubernetes?

What to coverThe posting names containerization, Kubernetes and Docker, and assigns ownership of deployment. Preparation suggestion: explain how you would package services, manage resources and roll out changes while limiting disruption.

What would you observe to know whether a model-backed scientific tool is working well?

What to coverThe engineer owns production observability for full-stack tools. Preparation suggestion: distinguish service-health signals from workflow outcomes, and explain how logs, metrics or traces could help locate a failure across the UI, API and model service.

How would you evaluate an LLM-assisted experiment-planning tool?

What to coverThe role includes systems that help propose experiments and learn from results. Preparation suggestion: address scientific review, uncertainty, traceability and whether suggestions improve the planning workflow; avoid assuming that an LLM's proposal is experimentally valid.

How would you connect experimental results back to model development?

What to coverThe posting emphasizes tighter feedback loops and reports measuring one million drug designs each month. Preparation suggestion: discuss linking designs to measured results, checking data quality and making results usable for analysis or subsequent training without assuming a particular data architecture.

How would you take a scientist-facing prototype into production?

What to coverThe role owns full-stack delivery from prototype through deployment and observability. Preparation suggestion: explain which assumptions you would test with users first, then what you would strengthen for reliability, maintainability and wider use.

When would you use AI development tooling while building this platform?

What to coverThe posting asks the engineer to use modern AI development tooling to move quickly while shipping quality work. Preparation suggestion: discuss tasks where it could speed implementation and how you would verify generated code, tests and behavior before release.

Behavioural practice

Tell me about a time you worked with a domain expert to uncover the real problem behind a tool request.

What to coverThis role requires close collaboration with wet-lab scientists and AI researchers. Preparation suggestion: use situation, action and result to show how you clarified the workflow, checked your understanding and adjusted what you built.

Tell me about a time you chose what to build first amid competing user needs.

What to coverThe role calls for product judgment and tools that support scientists, collaborators and partners. Preparation suggestion: describe the situation, the trade-offs you made, whose input you sought and the result, without implying a specific prioritization process at Nabla Bio.

Tell me about a time you owned a project from an early prototype into production.

What to coverEnd-to-end ownership is central to the advertised role. Preparation suggestion: explain your actions across implementation, deployment and follow-up, then state the result and what you learned.

Tell me about a time you shipped quickly without losing quality.

What to coverThe posting seeks high agency and a record of shipping quickly with quality. Preparation suggestion: give a real situation, the safeguards and compromises you chose, and the observed result rather than a general claim about working fast.

Tell me about a time feedback from users changed your technical approach.

What to coverThe role aims to tighten feedback between engineering, AI research and experimental work. Preparation suggestion: describe the original assumption, the feedback you received, your response and the result using a concrete situation/action/result example.

Sharpen your knowledge

Topics to prepare

Production model-serving architecture
Sketch how a drug-design UI, API and scalable backend service would expose a model. Explain how you would handle slow inference and failures. Relevant because the role explicitly requires experience serving ML models in production.
Training and inference system design
Draw the data flow from an experimental result to model development and a new inference request; identify where runs and outputs need to remain traceable. Relevant because the role builds robust training and inference systems.
Model–experiment feedback loops
Map a hypothetical cycle from proposed design to wet-lab measurement, analysis and the next design. Mark where an LLM might assist experiment planning and where scientists would review its output. This is a preparation exercise, not a specified Nabla Bio workflow.
Prototype-to-production trade-offs
Compare a quick scientist-facing prototype with a production release, naming what you would add for deployment, observability and reliability. Relevant because the engineer owns full-stack delivery through production.
Product judgment with scientists
Practice turning a wet-lab scientist’s reported bottleneck into a narrow first release and a way to check whether it helps. Relevant because the posting emphasizes understanding scientists’ pain points and shipping useful tools quickly.

Skills to revise

Python model-serving API
Build a small Python API around a placeholder model that accepts a design request, returns a prediction and handles invalid input. This rehearses the required Python fluency, API development and production model-serving work.
React and TypeScript dashboard
Create a small typed interface that submits requests to that API and displays results, loading states and errors. Relevant to the explicitly requested frontend skills and user-facing dashboards.
Training-to-inference pipeline
Implement a small practice pipeline using sample data that produces a model artifact and uses it for inference; document how an experimental result could feed the next run. Relevant to the advertised training and inference pipeline responsibilities.
Docker and Kubernetes deployment
Containerize the practice API with Docker and write a minimal Kubernetes deployment configuration; explain how you would update or roll it back. Relevant to the named containerization and cloud-native tools; this is a suggested drill, not a stated deployment setup.
Deployment observability
Add request, error and inference-duration signals to the practice service, then describe what you would inspect after a release. Relevant because deployment and observability are explicit parts of full-stack ownership.

Your preparation checklist

Map the qualifications to evidence
Prepare a one-page table of examples, where available, covering production ML model serving, React and TypeScript, Python, Docker and Kubernetes, and full-stack or ML infrastructure work. The posting asks for 5+ years in a relevant engineering role.
Rehearse a production model-serving walkthrough
Choose a real project, if available, and explain the backend architecture, inference path, scaling decisions, trade-offs and your contribution. Production systems that serve ML models are an explicit qualification.
Sketch a drug-design workflow tool
As a practice exercise, draw a UI, dashboard and API flow that could help scientists move from a design idea to an experiment. Explain how you would choose an initial scope; the posting calls for user-facing tools and product judgment.
Diagram the model-to-experiment feedback loop
Prepare a design exercise showing how training and inference pipelines could support experiment planning, analysis and learning from results. Label assumptions rather than presenting the design as Nabla Bio’s existing system.
Explain prototype-to-production delivery
Prepare an example, if available, of shipping quickly without sacrificing quality. Include deployment, observability, an issue you encountered and how you resolved it; the role owns delivery from prototype through production.
Prepare a scientific collaboration story
Identify a real example, if available, of translating a specialist’s pain point into a useful tool. Be ready to explain how you validated the need; for this role, the specialists include wet-lab scientists and AI researchers.
Draft questions about the role’s priorities
Bring questions about scientists’ current workflow bottlenecks, how LLM-assisted experiment proposals are evaluated, which tools need production work first, and how the team measures shorter design cycles.

Availability last checked: Nov 4, 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in United States. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…