Senior Full-Stack Engineer, AI/RAG Systems (Contract)

StrongAfter is hiring a contract Senior Full-Stack Engineer to lead hands-on engineering for OSWALT, our confidential, trauma-informed AI assistant for men who have had unwanted or abusive sexual experiences and the people who care about them. Partnering closely with our Product Owner and Program Manager, you'll turn priorities and feedback into reliable, well-tested improvements and help move OSWALT through its next stage of beta refinement. This is an AI-native engineering role: you pair disciplined use of coding agents with independent senior judgment, rigorous verification, and clear technical ownership. Six-month contract with potential to renew. ============================================================ ABOUT STRONGAFTER ============================================================ StrongAfter is a nonprofit working to make healing safer, more accessible, and less stigmatized for men who have had unwanted or abusive sexual experiences, and for those who support them. We pair trauma-informed expertise with carefully chosen technology to meet people where they are: privately, at their own pace, and without judgment. Our work is shaped by survivors, clinicians, advocates, nonprofit leaders, and product and technology specialists, and it reaches communities too often overlooked, including transitional-age young men, military and veteran communities, BIPOC men, and justice-involved men. Alongside OSWALT, we curate a Resource Hub of expert-vetted books, articles, and strength-building tools. Everything we build starts from one belief: that people deserve to explore what happened to them with dignity, choice, and hope. ============================================================ ABOUT OSWALT ============================================================ OSWALT is StrongAfter's AI assistant and content engine. It helps people ask hard questions and find trustworthy, trauma-informed information, drawn only from a curated library of expert-approved materials, never the open web. OSWALT is built for an early, often difficult moment: when someone is trying to understand what they're experiencing but isn't ready to talk to another person yet. Rather than improvising like a general chatbot, it matches a question to reviewed themes, retrieves approved summaries and excerpts, and composes a grounded, cited answer, which keeps its behavior predictable, inspectable, and possible to evaluate with care. OSWALT supports, but never replaces, professional care, crisis services, and human relationships, and it's currently in beta as we refine it and grow responsibly. ============================================================ THE OPPORTUNITY ============================================================ This is a high-ownership role for a senior engineer who enjoys working across a complete product in a sensitive domain. You'll guide OSWALT's ongoing engineering, turning priorities and feedback into thoughtful, well-tested improvements that raise reliability, answer quality, maintainability, and user experience. You'll partner closely with StrongAfter's Product Owner and Program Manager, and with trauma-informed subject-matter experts, including therapists, and product testers who evaluate OSWALT's responses, safety, tone, and user experience, working from concrete, reviewable artifacts and using AI coding agents throughout while staying independently accountable for everything that ships. ============================================================ WHAT YOU'LL DO ============================================================ - Own OSWALT's Python/FastAPI backend and Angular/TypeScript frontend, building, testing, and shipping updates end to end. - Use AI coding agents and LLMs across exploration, investigation, coding, tests, analysis, and documentation, directed by explicit requirements, acceptance criteria, and evaluation cases, and independently review and validate every change before it ships. - Partner with the Product Owner and Program Manager to refine requirements, set acceptance criteria, estimate and sequence work, and keep progress, dependencies, and tradeoffs visible. - Turn feedback from therapists, subject-matter experts, and testers into reproducible findings, test cases, and measurable evaluation criteria, distinguishing defects from preferences or content questions. - Improve answer quality through evaluation (embeddings, retrieval, reranking, grounding) against labeled tests and replay tooling, and maintain the content pipeline plus the knowledge-graph and vector retrieval behind cited responses. - Operate, troubleshoot, and improve OSWALT in its organization-managed GCP environment (NVIDIA L4 GPU capacity, containers, model serving, CI/CD, and observability), and protect grounding, citations, crisis signposting, and privacy-respecting design. - Produce right-sized, reviewable artifacts (issues, plans, code, tests, results, runbooks, release notes) and keep technical decisions and product-behavior documentation current. ============================================================ WHAT YOU BRING ============================================================ - Strong Python backend development, ideally with FastAPI or a comparable async framework. - Experience maintaining and shipping a production TypeScript application, and working productively in Angular. - Applied understanding of retrieval-augmented generation (embeddings, retrieval, reranking, grounding) and how to evaluate and improve answer quality. - Experience operating and troubleshooting cloud-hosted, containerized applications: Linux, Docker, and CI/CD. - Demonstrated fluency with AI-assisted, AI-native software development: you direct coding agents with explicit context and constraints, contain scope, critically review generated diffs, catch unsupported assumptions and unnecessary complexity, and validate behavior through tests and other evidence. You can independently understand, debug, and explain everything you ship, and you know when an AI tool is wrong. - An artifact-driven working style: you turn ambiguity into concise problem statements, requirements, acceptance criteria, plans, tests, evaluation results, and decision records that let product, program, engineering, and subject-matter experts collaborate effectively. - Senior engineering judgment in practice: designing and debugging independently, explaining tradeoffs, judging test quality rather than just whether tests pass, and keeping scope under control. - Security and privacy judgment for sensitive user interactions, and the ability to synthesize input from the Product Owner, Program Manager, subject-matter experts, and testers, reconcile conflicting needs, and communicate reasoning clearly, including through durable artifacts. We value demonstrated capability; adjacent tools and frameworks are welcome if you can show how they transfer. ============================================================ PREFERRED EXPERIENCE ============================================================ Any of these is a plus; none are required: - Google Cloud Platform. - NVIDIA L4 (or other) GPU inference, and vLLM or comparable model-serving systems. - Neo4j/Cypher, or other graph and vector retrieval. - Evaluation-driven improvement of retrieval and reranking. - Building AI products in safety- or privacy-sensitive settings. - Work in trauma, mental health, justice, nonprofit, or humanitarian technology. ============================================================ HOW WE WORK ============================================================ OSWALT is built through close collaboration among engineering, product, program, and trauma-informed expertise, working from concrete, reviewable artifacts rather than decisions that live only in chat threads or an agent's context window. The Product Owner guides direction and priorities; the Program Manager coordinates planning, delivery, and stakeholder feedback; and you own the technical approach, implementation, testing, and operational quality. Priorities become requirements and acceptance criteria; feedback from therapists, subject-matter experts, and product testers becomes reproducible findings and evaluation cases; technical decisions are captured concisely; and a finished change carries the code, tests, results, and notes others need to verify it. AI-assisted development is a requirement here, not an add-on. You'll use coding agents and LLMs throughout the workflow, kept on task by those same artifacts: the scope, requirements, and acceptance criteria that define what "done" means. Human accountability is nondelegable: you review and understand every proposed change and own architecture, correctness, security, privacy, safety implications, and what ships. Artifacts stay proportional to risk. A routine fix may need only a clear issue, a focused diff, tests, and a short note; larger changes add a written plan and explicit tradeoffs; and anything touching safety, privacy, crisis behavior, or grounding needs stronger evidence and stakeholder review. You'll engage thoughtfully with feedback rather than implementing every suggestion verbatim, clarifying the underlying need, weighing implications, and proposing options. Clinically consequential decisions (therapeutic content, crisis policy) stay with StrongAfter's qualified stakeholders; you implement and test approved requirements, protect grounding, crisis-signposting, and privacy behaviors, and close the loop once a change is validated. ============================================================ CONTRACT DETAILS ============================================================ - Engagement: Six-month contract with potential to renew - Hours: 10-20 Hours Per Week - Time zone overlap: U.S. Central Time - Start date: August 24th, 2026 ============================================================ HIRING PROCESS ============================================================ We hire with clarity and respect, starting with a brief intro call and assessing candidates through technical discussion, work samples, and references. Finalists may be invited to a paid, time-boxed trial of up to 10 hours. A trial is a likely step but not guaranteed; if we use one, it reflects the work the role involves and runs in a controlled environment with fixture data, no production access, and no sensitive user information. It is paid work, never unpaid production work. Because AI-assisted development is central to the role, a trial (or an equivalent technical discussion) is where you show how you work with these tools. You'll use AI as you normally would and give a concise record of material assistance: what you delegated, accepted, or rejected, and how you verified the result. You should be able to explain and defend every change independently. We evaluate independent technical reasoning, how well you direct and correct AI tools, scope control, code and test review, verification and artifact quality, connecting requirements to implementation and evidence, safety and privacy judgment, feedback synthesis, and clear communication, including explaining decisions without leaning on an AI tool. A final conversation with our top candidate precedes an offer. ============================================================ WHY STRONGAFTER ============================================================ This is a chance to do meaningful engineering: careful, high-ownership work on a product that helps people when trustworthy, judgment-free support is hard to find. If building something dependable, thoughtful, and safe appeals to you, we'd love to hear from you. Safe, seen, strong: what we want for the people we serve, and for the team building alongside you. ============================================================