Forward Deployed Engineer

Forward Deployed Scientist - Pharma

Location: Hybrid, regular on-site in Oxford, UK
Level: Senior / Principal

The role

We build AI systems that accelerate complex generics and 505(b)(2) formulation work, plus a causal engine that predicts what happens when you change a formulation, targeting PK and bioequivalence. It's a concierge model: clients never touch the platform, our scientists do.

You own the client relationship, run real discovery, judge whether our output is actually useful, and turn field constraints into requirements engineers can build.

What you'll do

Own the client relationship. Find out what's been tried, what failed, what's frozen, what the process can really do, and what decision they need to make. Half the job is discovery, before any analysis starts.

Gatekeep quality. Every deliverable goes through you first. Is it implementable at scale? Does it answer the real question? You can kill work before it ships.

Push past the obvious. If the output reads like a literature review, send it back. Force the causal engine and the client's own data to actually earn their keep.

Turn constraints into product requirements. "This process can't hit half a percent error" isn't small talk, it's a spec. Write it up so engineering can build against it.

Bridge both sides. Technical enough to challenge the engineering team, credible enough that a client's development lead treats you as a peer.

Compound every engagement. Log what was asked, what we found, what the data couldn't answer, and what the client valued. That becomes the evidence base.

Who you are

An industrial formulation / CMC scientist who has shipped real dosage forms and sat through a failed study.

Core experience

• PhD in pharmaceutics, pharma sciences, chemical engineering (or a Master's with more industry time)
• 5+ years in generics, complex generics, or 505(b)(2)
• Hands-on with solid oral dosage forms: modified/extended release, coating, dissolution methods, scale-up
• You've had a BE study fail, diagnosed why, and fixed it. Worth more than any credential
• You know what a plant can and can't reliably do: batch variability, yield loss, where lab-scale ideas break
• Fluent in FDA guidance, IID limits, ICH stability, PK endpoints, BE stats
• You read the data yourself and you'll get fluent in our stack. No need to code

AI fluency

• You already use AI tools daily, literature, data, protocols, reasoning, and have a real opinion on where they help and where they don't
• You've been burned by a confident wrong answer in your own field, and you can tell the story. That earned scepticism matters more than enthusiasm
• You interrogate AI output instead of accepting it
• When a tool falls short, your instinct is to fix it, not just complain

Also required

• You find the constraint everyone else assumed away
• Generalist range: mechanism, stats, process, regulatory, without handing off to a specialist each time
• You can tell a development lead their approach is broken and they'll thank you for it
• Intellectual honesty: "this data can't answer that, here's the experiment that would," beats a confident answer that fits the slide
• Comfortable with ambiguity, five people, no playbook, real deadlines