Application Architect - DARPA SciFy Project - Penn Engineering
Job Details
Application Architect - DARPA SciFy Project - Penn Engineering
University Overview
The University of Pennsylvania, the largest private employer in Philadelphia, is a world-renowned leader in education, research, and innovation. This historic, Ivy League school consistently ranks among the top 10 universities in the annual U.S. News & World Report survey. Penn has 12 highly-regarded schools that provide opportunities for undergraduate, graduate and continuing education, all influenced by Penn's distinctive interdisciplinary approach to scholarship and learning. As an employer Penn has been ranked nationally on many occasions with the most recent award from Forbes who named Penn one of America's Best Large Employers in 2023.
Penn offers a unique working environment within the city of Philadelphia. The University is situated on a beautiful urban campus, with easy access to a range of educational, cultural, and recreational activities. With its historical significance and landmarks, lively cultural offerings, and wide variety of atmospheres, Philadelphia is the perfect place to call home for work and play.
The University offers a competitive benefits package that includes excellent healthcare and tuition benefits for employees and their families, generous retirement benefits, a wide variety of professional development opportunities, supportive work and family benefits, a wealth of health and wellness programs and resources, and much more.
Posted Job Title
Application Architect - DARPA SciFy Project - Penn Engineering
Job Profile Title
Application Architect
Job Description Summary
We are seeking an Application Architect with expertise in NLP and large language models to lead development on the DARPA SciFy project. The University of Pennsylvania leads a multi-university team building AI systems for automated scientific claim assessment and technology forecasting, in partnership with Two Six Technologies. The architect will design and build the core infrastructure for ClaimSpy, our flagship claim feasibility assessment system, extending it from verifiable materials science claims to speculative AI technology forecasting - supporting DARPA's "create and avoid surprise" mission.
This is a 12-month grant funded position with the possibility of an extension dependent on continued funding.
Job Description
Job Responsibilities
- Design and build ClaimSpy's agent architecture: multi-LLM pipelines with tool-calling chains for literature retrieval, computational analysis, and claim assessment
- Develop an interactive Claim Card interface for analysts to construct, refine, and explore structured speculative claims, inspect reasoning traces, and review evidence bundles
- Extend the Panel of Virtual Experts (PoVE) system to support Tech Council in a Box (TCBX), enabling multi-perspective deliberation over speculative AI forecasts
- Integrate Two Six Technologies' speculative claims component into the ClaimSpy pipeline, including ablation studies to assess integration quality
- Instrument systems with logging, observability, and error tracing across tool-calling chains to support operational pilot deployments
- Collaborate with researchers on AutoRubric integration for automated LLM-as-judge evaluation of claim assessments
- Conform to DARPA's testing and evaluation framework
- Collaborate with postdocs and graduate students to translate research prototypes into production-ready systems
- Assist with technical documentation, system testing, and monthly DARPA reporting
- Participate in program meetings, technical demonstrations, and pilot exercises with transition partners
Qualifications
- Bachelor's in Computer Science with four to six years of substantial software development experience or an equivalent combination of education and experience.
- Strong Python skills and hands-on experience with LLM APIs (OpenAI, Anthropic, Google) and agent frameworks
- Experience building agentic systems with tool use, multi-step reasoning, and orchestration
- Experience developing APIs that expose LLM capabilities as services (REST/GraphQL endpoints, streaming, authentication, rate limiting)
- Experience with API development, system integration, and full-stack web development
- Ability to rapidly prototype and iterate in an agile research environment
- Strong communication skills for coordinating across a distributed team and with government stakeholders
Preferred
- Experience with RAG systems, vector databases, and retrieval pipelines
- Experience building interactive data exploration or sensemaking interfaces
- Familiarity with scientific literature processing and structured knowledge extraction
- Containerization and deployment (Docker, Kubernetes)
- Open source contributions
- Experience with evaluation frameworks, benchmarking, or LLM-as-judge pipelines
- Prior experience on government-funded research projects
Job Location - City, State
Philadelphia, Pennsylvania
Department / School
School of Engineering and Applied Science
Pay Range
$100,000.00 - $150,000.00 Annual Rate
Salary offers are made based on the candidate's qualifications, experience, skills, and education as they directly relate to the requirements of the position, and in alignment with salary ranges based on external market data for the job's level. Internal organization and peer data at Penn are also considered.
Equal Opportunity Statement
The University of Pennsylvania is an equal opportunity employer. Candidates are considered for employment without regard to race, color, sex, sexual orientation, religion, creed, national origin (including shared ancestry or ethnic characteristics), citizenship status, age, disability, veteran status or any class protected under applicable federal, state or local law.
Special Requirements
Background checks may be required after a conditional job offer is made. Consideration of the background check will be tailored to the requirements of the job.
Unlock this job opportunity
View more options below
View full job details
See the complete job description, requirements, and application process
