Artificial Intelligence Researcher

Kelly; San Francisco (CA); Hybrid; USD 145,000 - 170,000; Full time

Benefits offered by this job

Hybrid work arrangement

Job summary

Kelly Science & Clinical is seeking an AI Research Scientist for a direct hire in a biotech startup in the San Francisco Bay Area. The role is hybrid in SF, focusing on building AI benchmarks for therapeutic antibody development and working with leading AI and pharmaceutical teams.

The candidate will turn biologics development decisions into rigorous benchmark tasks, evaluate readouts, and collaborate with software engineers to create measurable evaluation standards in developability, PK,

Qualifications

• PhD in protein engineering, biochemistry, analytical chemistry, pharmacology, or related field.
• Hands-on experience analyzing antibody development data.
• Deep expertise in analytical or CMC characterization, large-molecule DMPK, protein engineering or developability, or nonclinical safety.
• Strong understanding of how experimental readouts relate to critical quality attributes, PK and effector properties.
• Python or R skills for data analysis.
• Regular use of AI tools and ability to judge model reasoning.

Responsibilities

• Turn development workflows into rigorous benchmark tasks.
• Develop evaluations across developability triage, analytical and CMC characterization, PK and FcRn reasoning, immunogenicity and safety, and first-in-human dose projection.
• Write gold-standard answers from raw data and set tolerances for variability.
• Curate data fixtures spanning developability panels and various analytical assays.
• Partner with software engineers to translate scientific judgment into practical workflows.

Skills

Python for data analysis

AI tools usage

Communicating uncertainty

Biologics data analysis

Developability knowledge

Education

PhD in protein engineering/biochemistry/analytical chemistry/pharmacology or related field

Tools

Python

R

Job description

Kelly® Science & Clinical is seeking an AI Research Scientist for a direct hire opportunity with one of our clients, a technology start-up in the San Francisco Bay Area market that builds AI agents to inform therapeutic antibody development. If you are driven by the prospect of translating cutting-edge scientific discoveries into tangible advancements and are poised to advance your career, join us as we pioneer progress in the biotechnology industry.

Workplace: Hybrid in the San Francisco, CA area

Position Title: AI Research Scientist, Biologics

Base salary range: 145,000-170,000/ year

Overview

Our client is building rigorous scientific benchmarks to evaluate whether AI agents can handle the complex, evidence-driven workflows used across biology and drug development. Working with leading AI and pharmaceutical teams, we test whether agents can reason through the messy, multi‑layered problems scientists face in practice, not just produce plausible answers.

We’re looking for an AI Research Scientist to help advance our benchmark for therapeutic antibody development. You’ll bring hands‑on biologics expertise to the work: turning real development decisions into challenging evaluations, establishing defensible scientific ground truth, and identifying where AI reasoning falls short.

The benchmark focuses on the period after lead nomination and before IND, spanning monoclonal antibodies, Fc‑fusions, and bispecifics. In this setting, agents must interpret analytical and preclinical data, distinguish genuine quality‑attribute risks from artifacts, and make advance, engineer, or stop recommendations that can withstand expert scrutiny.

You’ll work closely with scientists and software engineers to build the tasks, datasets, and evaluation standards that make those capabilities measurable.

Responsibilities

Turn development workflows into rigorous benchmark tasks

• Translate biologics development decisions into file‑based tasks with structured outputs and deterministic grading.
• Develop evaluations across developability triage, analytical and CMC characterization, PK and FcRn reasoning, immunogenicity, effector function and safety, and first‑in‑human dose projection.
• Write gold‑standard answers from raw data, setting tolerances that account for realistic variability in replicates, integration, and model fitting.
• Curate data fixtures spanning developability panels, glycan LC‑MS, charge‑variant CE, intact and subunit MS, SEC-MALS, FcRn and FcγR sensorgrams, concentration–time PK, and immunogenicity and effector assays.
• Partner with software engineers to translate scientific judgment into practical workflows and evaluations.

Set the standard for difficult scientific calls

• Assess whether an agent can distinguish a true liability from an analytical artifact—for example, a co‑eluting glycan peak, a misassigned charge variant, or an isolated assay flag not supported by orthogonal readouts.
• Evaluate whether agents interpret affinity and effector function in the context of a molecule’s mechanism, rather than assuming that “more is better.”
• Test whether agents can distinguish linear from target‑mediated clearance and assess a molecule against its target product profile.

Evaluate how agents reason

• Probe agent reasoning step by step and challenge conclusions that sound plausible but would not hold up in a development review or go/no‑go decision.
• Identify where and why an agent fails, what an expert would do instead, and what evidence could change the conclusion.
• Assess confidence calibration, rewarding explicit uncertainty when the evidence is incomplete over unjustified certainty.

Qualifications

• PhD in protein engineering, biochemistry, analytical chemistry, pharmacology, or a related field.
• Hands‑on experience analyzing antibody development data—you have performed the analysis yourself, not only overseen it.
• Deep expertise in at least one of the following: analytical or CMC characterization, large‑molecule DMPK, protein engineering and developability, or nonclinical safety; and working fluency across the others.
• Strong understanding of how experimental readouts relate to critical quality attributes, PK and effector properties, and release or comparability decisions.
• Python or R skills for data analysis, such as parsing chromatograms or electropherograms, fitting kinetics, or modeling PK.
• Regular use of AI tools such as Claude, Cursor, or similar agents—and the scientific judgment to recognize when technically coherent model reasoning is scientifically unsound.
• Confidence to challenge an AI agent when the evidence warrants it, and openness to revise your own view when presented with better evidence.
• Ability to communicate uncertainty and calibrate confidence rather than defaulting to binary conclusions.