
Forward Deployed AI Engineer, Delivery and Operations
Posted 16 days ago
The role in short
Embed with Delivery and Operations teams to redesign survey workflows by building, deploying, and maintaining production AI agents for vendor engagement, survey creation, and fielding.
About Sermo
Sermo runs the world's largest verified healthcare professional community, over 1.5 million triple-verified physicians, APPs, pharmacists and other clinicians across roughly 150 countries, and turns that community into real-time research and data products for pharmaceutical, biotech and medical device clients.
The role
You'll embed with Sermo's Delivery and Operations team and redesign the survey lifecycle around AI agents: from turning client requirements into a finished survey, to fielding it to the right healthcare professionals, to bringing in external vendors when more responses are needed. This is hands-on. You'll learn how the work runs today, then design, build, test, deploy and improve agent-first workflows with Delivery, Operations and Engineering. The goal isn't AI bolted onto today's process, but a redesigned workflow where agents do the work and people verify it.
Initial workstreams
Vendor engagement, the first major agent. When internal feasibility or response volume falls short, the team sources additional respondents from external market-research vendors through manual outreach, negotiation and tracking. You'll build the agent that detects when vendor support is needed, sends approved outreach, manages follow-ups, organizes responses, supports negotiation within the agreed rate card, tests survey links, learns from historical bid and pricing data, and escalates decisions that need human approval. Speed to answer is part of the value, since a faster quote wins more of the work.
Survey creation and translation. Client requirements arrive in many formats. You'll build a workflow that extracts and structures requirements from source documents, drafts the survey in internal tooling, flags unclear requirements and routes work to reviewers. You'll also improve AI-assisted translation, including selective re-translation of changed questions, with quality checks and review steps. You'll build on, or replace, the internal tools teams already use, bringing them within Sermo's data-handling and security standards.
Fielding and audience targeting. You'll improve how surveys reach the right healthcare professionals by specialty, country, screening criteria, exclusions and participation history; monitor responses; flag surveys at risk of missing volume or timelines; and recommend next actions, with fielding activity managed through Salesforce.
What you'll do
- Embed with Delivery and Operations to map the survey lifecycle end to end, including tools, handoffs, exceptions, approvals and bottlenecks, and set a measurable baseline.
- Redesign each workflow agent-first and build it: document extraction, structured outputs, routing, agentic task execution, and integrations through APIs, webhooks and automation tools.
- Connect workflows to Sermo's homegrown survey platforms and Salesforce, working with internal engineers on deeper platform changes.
- Design human-in-the-loop controls for every production workflow: approval checkpoints, review queues, guardrails, audit trails and exception handling.
- Build evaluations and monitoring, and track adoption, response rates, turnaround and time saved against the baseline.
- Document systems, workflow logic, prompts and operating procedures so the team can run and improve them.
- Identify and prioritize the next AI opportunities across Delivery.
What you'll bring
Required:
- 5+ years of software engineering, including 2+ years building and deploying AI-powered workflows, agents or automation in production business environments, not only prototypes.
- A completed agent-first redesign of a real business workflow that you can walk through: the baseline, the redesign, what people still do, and the measured result.
- Hands-on LLM experience with document extraction, classification, structured outputs, workflow routing and agentic task execution, plus evaluation of non-deterministic systems.
- Production engineering in Python and at least one typed language, with experience integrating CRMs, internal tools, APIs and data systems.
- Experience building human approvals, review queues, exception handling and quality controls into production systems.
- A proven ability to work directly with non-technical operators and turn complex workflows into tools they use, working from directional guidance and influencing stakeholders you don't manage.
- Strong written and spoken English, and professional Spanish fluency, since you'll work closely with Delivery stakeholders in Mexico.
- Full-time availability in Latin American time zones, collaborating with teams across North America and Europe, with occasional travel to the delivery office in Mexico.
Stretch, for an exceptional candidate:
- Several redesigns led across different functions, and the judgment to decide which workflows are worth doing next and build the business case for them.
- Experience architecting a shared agent platform, evaluation standards or guardrail patterns that other engineers reuse.
- A track record of setting direction across teams and influencing senior leaders.
- Experience mentoring engineers or helping build an applied AI team, and interest in growing into that role here.
Preferred:
- Salesforce, including workflows, data structures, outreach and activity tracking.
- C#/.NET, React and TypeScript environments, and connecting AI workflows to homegrown platforms.
- Healthcare, market research, survey operations or other regulated, data-sensitive environments.
- Multilingual or translation workflows.
- Zapier or similar automation platforms, and knowing when a purpose-built agent is the better choice.
- LLM observability, prompt and version management, and reliability testing.
Our stack
Homegrown survey tooling on C#/.NET, React and TypeScript, running on AWS; Salesforce and Zapier; Amazon Bedrock and AgentCore, Claude models and MCP for AI; Redshift and Databricks for data.
Your first 90 days
- 30 days: time on site with the Delivery teams in US and Mexico; relationships built with Delivery, Operations and Engineering; the vendor-engagement, survey-creation and fielding workflows mapped with baselines, decision points and required approvals; a phased plan agreed, with vendor engagement as the first pilot.
- 60 days: the vendor-engagement agent running in shadow mode on real work, with integrations, approval logic and monitoring in place and every output reviewed, then refined with the workflow owners.
- 90 days: at least one workflow live in production and used daily by Delivery, showing measurable reduction in manual work or faster turnaround against its baseline, and a prioritized roadmap for survey creation and fielding.
Education
BS in Computer Science or a related field, or equivalent practical experience.
Disclaimer
We may use artificial intelligence (AI) tools to support parts of the hiring process, such as reviewing applications, analyzing resumes, or assessing responses. These tools assist our recruitment team but do not replace human judgment. Final hiring decisions are ultimately made by humans. If you would like more information about how your data is processed, please contact us.
Prep for this interview
5 questions likely to come up at Sermo, each matched to a line in this posting.
- 1Integrations and APIs
They wrote: “Redesign each workflow agent-first and build it: document extraction, structured outputs, routing, agentic task execution, and integrations through APIs, webhooks and automation tools.”
Walk me through connecting our product to a customer's internal system that has no documentation.
How to answer
Why they ask: This happens weekly in the job. They want a calm, step-by-step method.
- Find who owns the system and get read access to a test environment
- Watch real traffic or logs to learn the data shapes
- Build a thin adapter with logging around every call
- Agree on how failures get reported, and who fixes what
Tip: Talk about finding the owner of the system first. The code is the easy part.
- 2LLMs, RAG and agents
They wrote: “Hands-on LLM experience with document extraction, classification, structured outputs, workflow routing and agentic task execution, plus evaluation of non-deterministic systems.”
A customer wants a chatbot over their internal documents. How do you build it?
How to answer
Why they ask: RAG is the most common first project. They want the whole path, from messy files to answers people trust.
- Collect and clean the documents; handle permissions per user
- Split them into chunks, embed, and store with source links
- Retrieve, then answer with citations; say “I don't know” when retrieval is weak
- Build a test set of real questions with expected answers before launch
Tip: Spend more time on the data and on testing than on the model.
- 3Discovery and scoping
They wrote: “A proven ability to work directly with non-technical operators and turn complex workflows into tools they use, working from directional guidance and influencing stakeholders you don't manage.”
A customer says “we want AI in our claims process.” What do you ask in the first meeting?
How to answer
Why they ask: Most FDE work starts this vague. They want to hear you find the real problem before you pick a tool.
- Map the current process: who touches a claim, in what order, with which systems
- Find the slowest or most error-prone step and ask what it costs them
- Agree on one measurable goal (for example time per claim, or error rate)
- Ask what data you can access, and who signs off on security
Tip: Ask about the work, not the technology. Who does it today, how long it takes, what a mistake costs, and how they'd know it got better.
- 4Cloud, deployment and security
They wrote: “Homegrown survey tooling on C#/.NET, React and TypeScript, running on AWS; Salesforce and Zapier; Amazon Bedrock and AgentCore, Claude models and MCP for AI; Redshift and Databricks for data.”
The agent reads emails and web pages. How do you stop text in them from taking control of it?
How to answer
Why they ask: Prompt injection is the main security risk of agents that act. They want concrete defences.
- Keep instructions separate from the content the agent reads
- Give the agent only the tools and permissions the task needs
- Require a person to approve actions that send, pay or delete
- Check outputs and log every tool call for review
Tip: Assume some content is hostile and limit what damage it can do.
- 5Coding
They wrote: “5+ years of software engineering, including 2+ years building and deploying AI-powered workflows, agents or automation in production business environments, not only prototypes.”
How do you keep quality high when you ship fast at a customer site?
How to answer
Why they ask: Speed is the job, but broken code at a customer is costly. They want your guardrails.
- Tests for the parts that touch customer data or money
- Code review even when it's fast, ideally by someone on the product team
- Feature flags and an easy way to roll back
- Clean up the shortcuts in the next release, and write them down
Tip: Name the few checks you never skip, and what you skip on purpose.
Not the one? Get the next one by email
New roles land here every day. Pick what you want to hear about.
Want the job, not just the posting? Learn how FDEs work.
FDE Info is our free weekly newsletter for forward deployed engineers: tools, field notes from real deployments, and how FDE teams hire and pay.
Free. One email a week. Unsubscribe in one click.
