AI Pipeline Engineer - REMOTE
Company Description
AI Pipeline Engineer
Arbitration Sciences Limited — Full-Stack — Remote
About the Role
Arbitration Sciences Limited is building a first-of-its-kind AI-powered platform that automates behavioral and linguistic analysis of international arbitration hearings. The system produces psychologically-grounded assessments of arbitrator, lawyer, and witness decision-making, delivering actionable intelligence to legal teams during proceedings.
The platform runs two parallel analysis pipelines (witness and arbitrator intervention), each employing independent LLM scoring runs with reliability validation before report generation. The methodology draws on computational linguistics, behavioral science, and tribunal psychology and is currently being built and empirically validated.
We have a comprehensive methodology specification and prompt suite, and need a full-stack AI application engineer to design and build the production platform. You will be the primary engineer responsible for transforming a detailed multi-stage analytical pipeline into an intuitive, push-button GUI application accessible to professionals with no technical background.
What You Will Build
• Multi-model API orchestration — Manage multiple LLM calls (Claude, GPT, etc.) with request queuing, timeout handling, and cost optimization across parallel scoring runs.
• Parallel scoring with reliability validation — Run independent scoring jobs concurrently, validate inter-rater reliability, and implement consensus logic before surfacing results to users.
• Transcript processing and semantic chunking — Handle variable-length legal transcripts, implement intelligent context windowing, and ensure each analytical segment receives appropriate context.
• Statistical consensus selection — Build decision logic that selects consensus outputs based on reliability metrics and flags low-confidence results for review.
• Cross-day cumulative tracking — Maintain state across multi-day proceedings, track behavioral patterns over time, and ensure continuity across analytical sessions.
• Report and visualization generation — Transform analytical outputs into professional reports and visual dashboards that communicate findings clearly to legal professionals.
Required Skills
• Strong backend development — Python or Node.js experience building robust APIs, async job processing, error handling, and state management at scale.
• Frontend development — React, Vue, or similar framework experience. Ability to design intuitive step-by-step workflows for non-technical users.
• LLM API integration — Direct experience calling Claude, OpenAI, or similar APIs. Understanding of prompt engineering, token management, and model-specific behaviors.
• Document processing — Experience ingesting PDFs, transcripts, or unstructured text; parsing, chunking, and preparing inputs for LLM analysis.
• Statistical computing — Comfortable working with reliability metrics, inter-rater agreement, consensus algorithms, and confidence scoring logic.
Nice to Have
• Legal technology experience
• Behavioral science or NLP background
• DevOps capability
• Database or persistent state management
What We Offer
• Ownership of the full technical stack — you’re not handed code, you’re building it from methodology up.
• Deep domain work — rare opportunity to shape how AI is applied to high-stakes legal proceedings.
• Collaborative environment — work directly with behavioral scientists, lawyers, and domain experts.
• Remote, flexible work — based where you are, structured around your productivity.
How to Apply
Send your resume and a cover letter that answers the following prompt:
This role involves orchestrating multiple LLM scoring runs in parallel and validating their reliability before surfacing results to users. Walk us through how you’d approach the technical trade-offs between speed, cost, and confidence in the consensus output — and tell us what part of this problem excites you most.
Email your application to: [email protected]. APPLICANTS who do not answer the prompt will NOT be considered.
Apply tot his job
Apply To this Job
AI Pipeline Engineer
Arbitration Sciences Limited — Full-Stack — Remote
About the Role
Arbitration Sciences Limited is building a first-of-its-kind AI-powered platform that automates behavioral and linguistic analysis of international arbitration hearings. The system produces psychologically-grounded assessments of arbitrator, lawyer, and witness decision-making, delivering actionable intelligence to legal teams during proceedings.
The platform runs two parallel analysis pipelines (witness and arbitrator intervention), each employing independent LLM scoring runs with reliability validation before report generation. The methodology draws on computational linguistics, behavioral science, and tribunal psychology and is currently being built and empirically validated.
We have a comprehensive methodology specification and prompt suite, and need a full-stack AI application engineer to design and build the production platform. You will be the primary engineer responsible for transforming a detailed multi-stage analytical pipeline into an intuitive, push-button GUI application accessible to professionals with no technical background.
What You Will Build
• Multi-model API orchestration — Manage multiple LLM calls (Claude, GPT, etc.) with request queuing, timeout handling, and cost optimization across parallel scoring runs.
• Parallel scoring with reliability validation — Run independent scoring jobs concurrently, validate inter-rater reliability, and implement consensus logic before surfacing results to users.
• Transcript processing and semantic chunking — Handle variable-length legal transcripts, implement intelligent context windowing, and ensure each analytical segment receives appropriate context.
• Statistical consensus selection — Build decision logic that selects consensus outputs based on reliability metrics and flags low-confidence results for review.
• Cross-day cumulative tracking — Maintain state across multi-day proceedings, track behavioral patterns over time, and ensure continuity across analytical sessions.
• Report and visualization generation — Transform analytical outputs into professional reports and visual dashboards that communicate findings clearly to legal professionals.
Required Skills
• Strong backend development — Python or Node.js experience building robust APIs, async job processing, error handling, and state management at scale.
• Frontend development — React, Vue, or similar framework experience. Ability to design intuitive step-by-step workflows for non-technical users.
• LLM API integration — Direct experience calling Claude, OpenAI, or similar APIs. Understanding of prompt engineering, token management, and model-specific behaviors.
• Document processing — Experience ingesting PDFs, transcripts, or unstructured text; parsing, chunking, and preparing inputs for LLM analysis.
• Statistical computing — Comfortable working with reliability metrics, inter-rater agreement, consensus algorithms, and confidence scoring logic.
Nice to Have
• Legal technology experience
• Behavioral science or NLP background
• DevOps capability
• Database or persistent state management
What We Offer
• Ownership of the full technical stack — you’re not handed code, you’re building it from methodology up.
• Deep domain work — rare opportunity to shape how AI is applied to high-stakes legal proceedings.
• Collaborative environment — work directly with behavioral scientists, lawyers, and domain experts.
• Remote, flexible work — based where you are, structured around your productivity.
How to Apply
Send your resume and a cover letter that answers the following prompt:
This role involves orchestrating multiple LLM scoring runs in parallel and validating their reliability before surfacing results to users. Walk us through how you’d approach the technical trade-offs between speed, cost, and confidence in the consensus output — and tell us what part of this problem excites you most.
Email your application to: [email protected]. APPLICANTS who do not answer the prompt will NOT be considered.
Apply tot his job
Apply To this Job