Badge
Eli Lilly & Co logo

PR&D – Advisor/Architect – Data & Agentic AI

Eli Lilly & Co

On-site · Full-time

🔒 Unlocks at Trust Score 90🔒 Unlocks at 90

Job summary

Eli Lilly & Co is a global healthcare company focused on developing life-changing medicines. In this role, you will work on building and managing data pipelines, machine learning, and natural language processing components to support the Product Research & Development team. Key skills required include Python programming, data engineering, and experience with ML and AI systems. This position is suitable for someone with a strong technical background who enjoys hands-on work and collaboration. The pay is USD 175,000–8,000,000 a year. This job unlocks at Trust Score 90 on Badge.

About the job

At Lilly, the work is demanding because patients are waiting. We unite caring with discovery to help make life better for people around the world, knowing that every decision, every detail, and every day matters. Headquartered in Indianapolis, Indiana, our over 50,000 employees around the globe take on complex challenges to discover and deliver life-changing medicines, strengthen how health is understood and managed, and support the communities we serve. This is hard, urgent, selfless work—but it’s work worth doing. If you’re driven by purpose and ready to bring your best to work that truly matters for patients, we invite you to join us. 


Product Research & Development (PR&D) owns the CMC (Chemistry, Manufacturing, and Controls) data that underpins every regulatory submission, batch investigation, and process decision across Lilly's development portfolio, spanning structured systems (LIMS, MES, DMS, ELN) and unstructured scientific content (protocols, batch records, investigation reports). The function operates under strict GxP, data integrity, and audit-readiness requirements. The Advisor/Architect – Data & Agentic AI, PR&D, builds the engineering backbone that carries PR&D content from source to a governed, served agent: ingestion pipelines, ML and NLP components, retrieval services, APIs, and agent runtimes that sit on the Semantic Layer and Knowledge Graph. This role is hands-on, writing and shipping production code across data engineering, ML, and agentic delivery, and equally operates at design and guidance level: setting reference designs, reviewing team designs, and advising the teams that build on the platform. It works alongside the Systems Architect and knowledge graph engineers who own ontology and graph design. This role ensures PR&D's agents are built on shared, tested, reusable components rather than one-off pipelines, so each new agent and domain ships faster and stays governable as the footprint scales.

Path/Level: R4/R5 (Advisor/Architect), individual contributor.

Key Responsibilities

  • Multi-Modal Data Ingestion — build and operate ingestion pipelines for multi-modal and multi-lingual PR&D content (documents, structured records, and beyond text-only formats where relevant), feeding the knowledge graph and vector store with ontology-based tagging.
  • ML & NLP Components — build entity and relationship extraction, classification, embedding, and LLM-assisted metadata enrichment components, and productionize them with testing, versioning, and monitoring.
  • Retrieval Services — implement and tune hybrid graph, vector, and keyword retrieval services for human exploration and agentic tool calls, optimizing precision, recall, latency, and groundedness against agreed evaluation sets.
  • Agent Runtimes & MCP Services — build agent runtimes and MCP (Model Context Protocol) servers that expose Semantic Layer capabilities (Q&A, summarization, context retrieval) with tool calling, orchestration, and human-in-the-loop controls.
  • Evaluation, Observability & Guardrails — build evaluation harnesses, tracing, monitoring, and guardrails for agents, and contribute to a shared prompt library, so PR&D's agentic footprint stays measurable and governable beyond single-agent use cases.
  • Full-Stack & API Delivery — design and deliver APIs, services, and internal tooling (including front-end where needed) that make Semantic Layer and agent capabilities consumable by domain teams and scientists.
  • Platform Engineering & DevOps — own CI/CD, containerized deployment, environment management, and monitoring on Azure or AWS, with infrastructure and releases that meet GxP change-control expectations.
  • Data Quality, Lineage & Compliance Support — build data-quality checks, lineage, and access controls into pipelines and services, and produce the traceability and validation evidence needed to meet FDA 21 CFR Part 11, GxP, and enterprise security and privacy requirements.
  • Design Authority & Technical Guidance — act as design authority for ingestion, retrieval, and agent services: define reference designs, review team designs and code, advise domain teams on build-versus-reuse decisions and trade-offs, and give technical guidance to engineers building on the platform, while remaining a primary builder.
  • Reusable Engineering Patterns — establish reference implementations for ingestion, retrieval, and agent services that domain teams can reuse, reducing fragmentation and delivery time.
  • Technology Evaluation — evaluate vector database, orchestration, and agentic frameworks for CMC applicability through staged (Bronze/Silver/Gold) proofs of concept before wider adoption.
  • Collaboration & Peer Mentorship — work closely with the Systems Architect, knowledge graph engineers, and platform engineers; mentor peers on production data and agent engineering practice.

Technical & Strategic Expertise Areas

Python · Production Data Engineering (ETL/ELT, batch and streaming) · ML & NLP Engineering · Entity/Relationship Extraction · Embeddings & Vector Databases (Azure AI Search, AWS OpenSearch, Pinecone) · Hybrid & Graph-RAG Retrieval · Agentic AI & Multi-Agent Systems · Agent Orchestration (LangGraph or equivalent) · MCP (Model Context Protocol) · Agent Evaluation & Observability · Prompt & Context Engineering · API & Service Design (REST) · Full-Stack Development · Cloud Platforms (Azure/AWS) · Containerisation & Kubernetes · CI/CD & Infrastructure as Code · Data Quality, Lineage & Access Control · Knowledge Graph Fundamentals (Neo4j, RDF/OWL) · Reference Architecture & Design Review · Security & Compliance Awareness · Stakeholder Communication

Basic Qualifications

  • Proven track record delivering production data, ML, and agentic systems end to end, building them rather than overseeing others who build them.
  • Hands-on production Python across data engineering and ML.
  • Experience operating agents in production with tool calling, orchestration, evaluation, and observability, beyond single-agent RAG or Q&A systems.
  • Experience building retrieval systems on vector stores and tuning them against measured quality.
  • Evidence of end-to-end ownership, from ingestion through to a served, monitored endpoint.
  • Experience operating at both levels: building hands-on, and defining reference designs, reviewing others' work, and advising teams on trade-offs.
  • Experience partnering with architects, engineers, and business or science stakeholders to turn requirements into delivered services.

Additional Qualifications / Competencies

  • Exposure to Neo4j or GraphDB, RDF/OWL, and Graph-RAG patterns.
  • Experience in pharmaceutical, life sciences, or other GxP-regulated environments, including traceability, validation, and access control.
  • Experience exposing knowledge or data services to LLM and agent consumption via MCP or equivalent tool-calling interfaces.
  • Experience with multi-modal or multi-lingual content pipelines.
  • Front-end or internal-tooling experience.
  • Comfortable explaining engineering trade-offs to senior, non-technical stakeholders.

Required Leadership Capabilities

  • Goes deep technically and communicates trade-offs in business-relevant terms.
  • Gives design guidance and advice to teams building on the platform, and reviews designs and code, while remaining a primary builder.
  • Takes ownership of a component or capability and sees it through to production and operation.
  • Raises design concerns early and unblocks peers without needing a formal people-manager structure.
  • Systems thinking that connects pipeline, retrieval, and agent design choices to downstream scientific workflows and business outcomes.

Lilly is dedicated to helping individuals with disabilities to actively engage in the workforce, ensuring equal opportunities when vying for positions. If you require accommodation to submit a resume for a position at Lilly, please complete the accommodation request form (https://careers.lilly.com/us/en/workplace-accommodation) for further assistance. Please note this is for individuals to request an accommodation as part of the application process and any other correspondence will not receive a response.

Lilly does not discriminate on the basis of age, race, color, religion, gender, sexual orientation, gender identity, gender expression, national origin, protected veteran status, disability or any other legally protected status.

#WeAreLilly

Free · 2 minutes. On Badge, recruiters see reviews from people you've worked with, not just your CV.

How can Badge help me land interviews?

Recruiters get thousands of applications. Stand out with anonymous reviews from people you've worked with, and build a Trust Score that gets you more interview calls.

  1. 01

    Build your professional reputation

    Give and collect anonymous reviews with colleagues you have worked with. Each review adds to your Trust Score, and it stays yours for life.

  2. 02

    Showcase your Badge

    Put your Badge QR on your CV and the Badge banner on LinkedIn. Recruiters and clients scan it and see your verified reviews.

  3. 03

    Land interviews

    At 70, jobs on Badge open to you. A higher score ranks you above other applicants, so you get more interview calls.

Ask your teammates to review your work on Badge. It is free, and the score goes with you to your next job.

FAQs

Recruiters get thousands of applications, and most CVs look the same. Badge shows them anonymous reviews from people who actually worked with you, plus your Trust Score. On Badge jobs, applicants are ranked by Trust Score, so a higher score puts you at the top of the recruiter's list. That proof is what helps you stand out and get the interview call.

Getting your first review on Badge takes about a minute — you don't have to chase anyone or write awkward "please review me" messages. Here's how it works: 1. Install the Badge app and create your profile. 2. Give your first anonymous peer review to a colleague you've actually worked with — that's what starts your Trust Score. 3. Share your phonebook so Badge can find the colleagues who can review you back, and reach out on your behalf with short, targeted, anonymous review questions. 4. Their answers land directly on your profile and push your Trust Score up — which is what unlocks startup jobs.

Every job on Badge unlocks at a Trust Score. Entry jobs are open to everyone at the base score of 60; better-paying and more senior roles unlock higher — 75, 85, 95. You raise your score by giving reviews and completing the steps the app puts in front of you. Giving a good peer review is the biggest, most repeatable jump. When your score passes a role's threshold, the Apply button unlocks in the app. Your score is live — it's recalculated from your network, so it can move both ways. If it drops below a role's threshold you earn it back the same way: give more reviews, finish your profile.

Every time you change jobs, you start from zero. Your new team doesn't know you delivered that impossible deadline, that you mentored three juniors, or that your last manager called you the most reliable engineer they'd worked with. Years of hard work — invisible. And when layoffs hit, you're just another resume in a pile of AI-generated applications with fake endorsements. Badge fixes this. It collects anonymous peer reviews from colleagues you've actually worked with — your managers, teammates, and clients. Those reviews build your Trust Score. It's like a credit score, but for how you work. Badge is your proof of work — evidence that you actually did what your resume claims. The score travels with you. Change jobs, switch industries, get laid off, become a freelancer — your reputation follows. Because Badge reviews are verified and from real colleagues, your profile stands out against candidates padding their resumes with fake AI-generated reviews. Recruiters on Badge discover candidates by Trust Score, so a higher score means you show up first when opportunities open. You stop competing on resume keywords and start competing on what actually matters: what people who worked with you say about you.

LinkedIn endorsements are public and named — so people only write positive things. Recruiters and hiring managers already know this. Most endorsements come from friends or people returning the favour, not from colleagues giving an honest assessment of your work. Badge reviews are anonymous and specific. Our AI agents ask reviewers targeted questions about how you actually work — your reliability, collaboration, communication, how you handle pressure. Because it's anonymous, people share the real picture, not just polite compliments. It's your proof of work — verified by the people who were actually there. That specificity is what makes Badge valuable. Instead of generic "highly recommended!" endorsements, your profile shows detailed reviews from people who've worked with you, answering questions that actually matter to recruiters. Ten specific anonymous peer reviews carry more weight than fifty one-click LinkedIn endorsements from connections who barely remember working with you.

Your Trust Score is the key to the jobs board. Jobs are gated by score — the higher yours, the more jobs you can apply to — and inside a role, applicants are ranked by Trust Score, so a higher score also puts you at the top of the recruiter's list. Right now, you're invisible. You apply to jobs, your resume goes into an ATS, and a keyword filter decides your fate. Doesn't matter how good you are — if your resume doesn't match the right keywords, nobody sees you. Badge flips this. Recruiters on Badge browse and search candidates by Trust Score. Instead of you chasing jobs, recruiters find you. A higher score means you appear higher in recruiter searches — your verified peer reviews do the selling before you even know the opportunity exists. Here's how it works: recruiters search for candidates by role, industry, or company. Results are ranked by Trust Score. A candidate with 15 anonymous peer reviews and a strong score shows up before someone with a polished resume and zero verified reviews. Your proof of work puts you ahead of the crowd. The more reviews you collect over time, the stronger your score, the more visible you become. Your reputation compounds instead of resetting every time you switch jobs.

This is the first thing everyone worries about — what if someone who didn't like me leaves a bad review? It's a fair concern. The short answer: one bad review won't hurt you. Your Trust Score is built from all your reviews over time. If 14 colleagues say you're reliable, collaborative, and deliver on time, and 1 person says you're difficult — that one person looks like the outlier, not you. Volume protects you. Badge also has safeguards. Abusive or toxic reviews are automatically detected and made private. You can make any review private if you don't want it visible. You can report spam or fake reviews. The system isn't designed to let one person with a grudge define your career. And here's the thing most people don't consider — a few constructive critical reviews actually make your profile more credible. A profile with nothing but glowing praise looks suspicious. A profile that shows you're great at collaboration but still growing as a public speaker? That's a real person. Recruiters trust that more.

Fake reviews are impossible — every reviewer must prove a verified work overlap with you (same company, same time period) before the review is accepted. Reviews from verified colleagues carry more weight in your Trust Score than reviews from loose connections. The review process itself filters out low-quality responses. Our AI agents guide reviewers through targeted questions about your actual work — not generic "rate this person 1-5" prompts. If someone tries to leave a vague "great person!" review, it gets rejected. The system demands specifics. And if someone you've never worked with tries to leave a review? It carries minimal weight because there's no verified work overlap. You can also report reviews that are fake, spammy, or from people you've never interacted with. The system is built to reward genuine reviews, not let strangers or bad actors influence your score.

Yes — you can see every review on your profile. All reviews on Badge are anonymous. The review content is visible to you and recruiters, but the reviewer's identity is always hidden. That's by design. When people know their name won't be attached, they write what's actually true instead of what's socially safe. That honesty is what makes your Badge profile worth more than a LinkedIn page full of polite endorsements from people who didn't want to say no.

More jobs like this