for to design and develop next-generation, agentic AI tools that revolutionize complex document review and data analysis workflows. In this role, you will build intelligent multi-agent systems that allow users to interrogate dense technical documents, execute multi-step analytical tasks, and automatically populate operational dashboards.
You will orchestrate advanced generative AI, Retrieval-Augmented Generation (RAG), Model Context Protocol (MCP), and automated evaluation systems to create highly reliable, secure, and self-correcting software architectures. You will collaborate closely with data standardization specialists and ontologists to integrate advanced data schemas and controlled terminologies into the AI pipeline.
Design and implement working agentic AI prototypes using frameworks like LangGraph, LangChain, or CrewAI capable of executing multi-step analytical and reasoning processes for document reviews
·
Interactive Query Interfaces:
Create generative AI interfaces that allow users to query, interrogate, and extract granular insights from dense, structured, and unstructured documentation.
2. Intelligent Data Extraction & Query Pipelines
·
Query Augmentation:
Build pipelines that automatically inject external metadata and ontologies into LLM prompts to maximize query accuracy.
·
Dual-Stream Parsing:
Code high-precision extraction strategies to ingest and parse data from both modern structured formats and legacy unstructured documents.
·
Similarity Matching:
Develop algorithmic workflows to compare newly processed documents against historical databases using metadata clustering and vector similarity.
3. Pipeline Automation & System Integration
·
End-to-End Automation:
Build working agentic AI prototypes that autonomously extract/analyze technical information and use it to dynamically populate downstream user dashboards and analytical tools.
·
Connected Systems:
Utilize Model Context Protocol (MCP) to seamlessly connect LLM agents to internal data sources, external databases, and developer/user interfaces.
4. Validation, Metrics & Continuous Learning
·
Confidence Scoring:
Design and implement automated confidence scoring mechanisms and LLM-as-a-judge frameworks to estimate the accuracy of query results and proactively alert users when manual review is needed.
·
Feedback Loops:
Program feedback processes to capture user input and error patterns, enabling continuous model, prompt, and routing improvement.
·
Extensible Documentation:
Document architectural patterns, lessons learned, and framework constraints to allow the methodology to scale across other business units and regulatory review streams.
5. Data Security, Governance & Auditability
·
Audit Trail Architecture:
Implement comprehensive, stateful logging across all multi-agent steps, ensuring every data point extracted or populated into user dashboards can be traced back to its exact source snippet in the original documentation.
·
PII & Data Privacy Guardrails:
Design and embed automated preprocessing layers to detect, redact, or safely handle Personally Identifiable Information (PII) and sensitive corporate data before it is processed by external LLM APIs.
·
Enterprise Security Compliance:
Ensure all agentic pipelines, vector databases, and Model Context Protocol (MCP) integrations strictly adhere to enterprise data isolation, encryption-at-rest, and encryption-in-transit protocols.
·
Access Control & Permissions:
Implement secure role-based data routing within the agent logic, ensuring the AI system only retrieves and displays information that the querying user has explicit permission to view.
RequirementsRequired
Technical Skills:
·
Education:
Bachelor’s or Master’s degree in Computer Science, AI/ML, or equivalent practical experience.
·
Core Development:
Proficiency in
Python
and other languages commonly used in AI/Agentic development (such as
TypeScript/JavaScript
or
C++ ), backed by strong software engineering fundamentals (OOP, CI/CD, version control).
· Experience with Clinical trial protocol review or FDA is plus
·
LLM Application Development:
Hands-on experience building applications with LLM APIs (OpenAI, Anthropic, Google, etc.).
·
Agentic Frameworks & Architecture:
Deep understanding of RAG architectures, Model Context Protocols (MCPs), and agentic AI frameworks (e.g., LangGraph, Autogen, LangChain, ReAct/ReWOO).
·
Search & Semantic Data:
Experience with vector databases and embedding models for semantic search, as well as handling complex data parsing (structured vs. unstructured data).
·
Prompt Engineering:
Strong knowledge of advanced prompt engineering techniques, in-context learning, and optimization for dense text to minimize hallucinations.
·
MLOps & Cloud:
Experience with MLOps practices including model/prompt versioning, monitoring, and deployment, alongside familiarity with cloud platforms (AWS, GCP, Azure) and containerization (Docker, Kubernetes).
·
Validation & Evaluation:
Experience designing programmatic validation metrics, LLM-as-a-judge patterns, and confidence-scoring algorithms for AI outputs in high-stakes environments.
Core Professional Competencies:
·
Problem-Solving:
Excellent problem-solving skills and the ability to work effectively with ambiguous requirements.
·
Communication & Collaboration:
Strong communication skills to explain complex AI concepts to various stakeholders, and experience collaborating with Data Engineers, Ontologists, or Data Standardization specialists to turn data models into functional AI context.
·
AI Safety:
Solid understanding of AI safety, alignment, and ethical considerations.
Nice-to-Have (Preferred Qualifications):
·
Evaluation Ecosystems:
Experience with evaluation frameworks (e.g., Ragas, LangSmith) to benchmark agent performance on complex reasoning tasks.
·
ML Frameworks:
Proficiency with traditional ML frameworks (PyTorch, TensorFlow) and the Hugging Face ecosystem.
·
Emerging Tech:
Curiosity regarding emerging AI technologies complementary to generative AI (e.g., Graph RAG) to further improve system quality.
Benefits
PPO/HMO Health Plan (includes medical, dental, and vision)