# Syntheos > Syntheos designs and ships working platforms for federal R&D funders, defense leaders, research institutions, and universities. Every conclusion traces back to its source. Our systems run today at DARPA, the Andrew W. Marshall Foundation, and Georgetown. Syntheos is a strategic advisory and engineering firm that builds decision systems that survive an audit. Each engagement ends with a working system the client's team runs, not a slide deck. Founded by Caleb Smith. Based in the United States. The firm does not build a single product. It builds four shapes of working system, each grounded in a system already shipped to a real client. Where most vendors call their work "traceable" or "auditable," Syntheos can show what those words mean in code: evidence ledgers, verified-versus-inferred labeling, delegation contracts, and W3C PROV lineage. ## What Syntheos builds - [Program Analysis Products](https://syntheos.io/work/darpa-program-analysis): Fixed-scope analytical sites where every quantitative claim ships with a reproducible query and a confidence level. Built for R&D funders who have to defend portfolio decisions on the record. Reference engagement: DARPA program offices. - [Decision Platforms (Deepfield)](https://syntheos.io/work/deepfield): A nine-stage assessment pipeline with W3C PROV lineage on every output. Knowledge graph, MCTS reasoning, wargaming. Built for standing net-assessment and competitive-intelligence work. - [Research Intelligence Tools](https://syntheos.io/work/awmf-archive): Knowledge-graph research consoles over specialist corpora. Hybrid retrieval, chat with citation badges, and a visual distinction between verified archive material and AI-generated inference. Reference engagement: Andrew W. Marshall Foundation. - [Human-AI Teaming Systems](https://syntheos.io/work/georgetown-human-ai-teaming): Orchestrated AI assistance with human judgment in the loop. A delegation contract constrains what the AI is allowed to do, and database-level phase gates enforce it. Reference engagement: Georgetown SEST Wicked Problems Lab. ## How Syntheos shows its work - [Evidence ledgers](https://syntheos.io/how-we-show-our-work): Every quantitative claim on a Syntheos site is backed by a structured ledger entry with the claim, a confidence level (HIGH, MODERATE, LOW), and a reproducible source query. A discrepancy between the ledger and live data stops the build. - Verified versus inferred labeling: Knowledge-graph nodes are visually distinguished by source. Verified means the content came from the archive. Inferred means the AI generated it during reasoning. Citations resolve to specific passages in specific documents. - Delegation contracts: A small set of constraints written into code that the AI cannot verify assumptions, cannot define success criteria, and cannot advance past a human gate. Gates are PL/pgSQL stored procedures, enforced by the database, not by hope that the model will behave. - W3C PROV lineage: Every recommendation traces back through the entire reasoning chain to source evidence. Every module returns a typed Result, so failures propagate explicitly. Silent fallbacks are forbidden by the architecture. ## Past performance and reference clients - DARPA — analytical sites for program analysis at multiple program offices. - The Andrew W. Marshall Foundation — research intelligence console over the archive. - Georgetown University SEST — Wicked Problems Lab human-AI teaming platform. - Center for Security and Emerging Technology (CSET) — bibliometric analysis frameworks. - Noble Reach Foundation — research and analysis on the technology innovation pipeline. - University of Michigan Medical School, RTI International, the American Board of Anesthesiology, Digital Science. ## Industries served - [Defense and national security](https://syntheos.io/defense): competitive assessment, technology scouting, S&T assessment, decision architectures for complex operations. SAM.gov registered. NIST SP 800-171, DFARS 252.204-7012, and CMMC 2.0 Level 2 in progress. NAICS 541611, 541715, 541512, 541690. - Tech transfer and federal R&D portfolio decisions. - Policy and research funding. - Healthcare and biomedical research. - Infrastructure and public risk. ## Key pages - [Home](https://syntheos.io/) - [Work portfolio](https://syntheos.io/work) - [How we show our work](https://syntheos.io/how-we-show-our-work) - [Defense capabilities](https://syntheos.io/defense) - [Insights (long-form essays)](https://syntheos.io/insights) - [Publications (peer-reviewed papers)](https://syntheos.io/publications) - [About](https://syntheos.io/about) - [Contact](https://syntheos.io/contact) ## Selected insights Syntheos publishes long-form essays on AI decision-making, institutional risk, and the gap between what systems promise and what they actually do. Topics include sycophancy in RLHF systems, the calibration crisis in expert-AI advisory layers, the half-life of professional expertise, the inverse confidence law, and the taxonomy of organizational silence. See the full archive at https://syntheos.io/insights. ## How to cite Cite as: Syntheos, https://syntheos.io. Founder and CEO: Caleb Smith. Direct contact: info@syntheos.io. For defense and national security inquiries: defense@syntheos.io. ## Crawler policy Syntheos welcomes indexing by AI agents and search crawlers. The full sitemap is at https://syntheos.io/sitemap.xml. Robots policy at https://syntheos.io/robots.txt. The site is statically generated, so a polite single-pass crawl reflects the canonical content.