AI Safety Argumentation Platform Research Engineer

About CARMA: The Center for AI Risk Management & Alignment (CARMA) is a forward-thinking, mission-driven research institution working to safeguard humanity and the biosphere from transformative AI risks, including AGI and ASI cataclysms. Operating as a fiscally-sponsored project of Social & Environmental Entrepreneurs, Inc. (a 501(c)(3) public benefit non-profit), CARMA constructs rigorous analytic infrastructures, knowledge platforms, and robust policy frameworks to ground enterprise and global AI governance before outsized tail-risks materialize.

Position Overview

We are seeking an intellectually deep, systems-focused AI Safety Argumentation Platform Research Engineer to develop, operate, and scale our flagship evidentiary infrastructure. In this zero-to-one engineering track, you will build computational systems where formal argumentation theory meets autonomous agentic AI. You will extend graph ontologies representing complex safety claims, implement defeasible reasoning frameworks, and manage LLM-driven population pipelines. Serving as the technical backbone of our research arm, your machinery will translate non-deterministic safety arguments into tractably verified, persuasive content optimized for global policymakers, technical researchers, and journalists.

Key Responsibilities

• Knowledge Graph Architecture: Extend and maintain ontology schemas representing intricate claims, evidentiary weights, inference lines, and semantic defeaters using OWL/RDF structures.
• Computational Argumentation Modeling: Implement formal defeasible argumentation frameworks (e.g., ASPIC+ and Dung-style abstract topologies ) to capture logical rebuttals and counterargument dynamics.
• LLM Pipeline Operations: Build and quality-control automated data-harvesting pipelines, embedding cross-check verification scaffolds, strict provenance logs, and human-in-the-loop curation gates.
• Agent Coordination Design: Architect multi-step agent orchestration patterns to scrape, digest, and populate database nodes, incorporating advanced error handling and graceful degradation logic.
• Objection Steel-Manning: Pre-harden systemic argument frameworks by procedurally mapping adversarial counterclaims, cognitive bias indicators, and highly verified rebuttals.
• Multi-Format Export Workflows: Build automated pipeline endpoints that map graph data structures into diverse, audience-optimized communication formats across various registers.

Required Skills & Qualifications

• Proven technical familiarity with formal argumentation theory, abstract/structured computational dialogue, or defeasible reasoning models.
• Hands-on experience developing knowledge graphs, properties registries, or semantic frameworks utilizing graph databases and custom query languages.
• Operational history managing LLM agent systems, multi-agent frameworks, prompt engineering abstractions, and adversarial verification regimes (consistency checks, calibration).
• Active, highly agile “vibecoding” execution habits—adept at rapid software prototyping and deployment using AI-assisted engineering tools.
• Substantive contextual grounding in frontier-AI dynamics, technical safety philosophies (alignment, capabilities tracking), and global AI governance vectors.
• Familiarity with philosophy of science primitives bearing on evidence evaluation, including inference to the best explanation, underdetermination, and burden of proof concepts.
• Location Context: 100% remote-first operational infrastructure framework open to qualified technical professionals residing Anywhere Globally (Open Worldwide) .
• Employment Structural Note: United States-based hires will be onboarded as standard full-time W2 employees; international hires will be engaged via long-term independent contractor agreements. Visa sponsorship is unavailable.

Preferred Strategic Indicators (Nice to Have)

• Graduate degree or equivalent computational research history in Argument Mining, Epistemology, Computational Argumentation, or Philosophy of Science.
• Hands-on familiarity with specific argument representation frameworks such as AIF or Carneades .
• Track record of published safety papers, policy contributions, or open-source software contributions within the global AI alignment ecosystem.

What We Offer

• Targeted Annual Salary: $160,000 – $210,000 USD per annum (Calibrated accurately against research background, experience depth, and contract classification).
• The exceptional opportunity to engineer a world-class epistemic engine reshaping the landscape of global AI safety communications.
• Complete geographic remote flexibility paired with structured allowances for occasional international travel and company retreats.
• A highly collaborative, inclusive non-profit research team operating with the technical rigor of an cutting-edge AI startup.