About Us We blend best practices from outsourcing to deliver services that feel in‑house: low‑risk delivery with high quality, stable cross‑functional teams, and a business‑first approach. We invest in people, business development, and clear communication. Our domain expertise spans healthcare, automotive, education, fintech, and media/entertainment. For this role, you will work on Banner Copilot for Banner Aircraft International. The strategic goal is to turn Banner Copilot into an AI‑supported operating and decision‑making layer: automate standard account‑management and operational workflows, speed up transactions, surface risks, and provide data‑driven recommendations. The platform will start with decision support and next‑best‑action guidance, then progress to controlled automation in clearly defined, low‑risk scenarios.

The role Role purpose:

Deliver a production‑grade AI automation layer that reduces manual operational work, improves decision quality and speed, and highlights business and financial risks with concrete, data‑backed recommendations.

Work setup:

Remote; Part‑time,

Contract Responsibilities:

Product and solution design: Audit the current system and architecture Map key operational and management decisions; identify what to automate now vs later Define the transition from decision support to controlled automation and limited autonomy Set the target architecture for the RFQ → Quote → Sales Order → Delivery → Close → Follow‑up → Management Recommendation cycle Separate business rules, algorithms/statistics, ML models, and AI agents Create a phased technical roadmap from the current solution to an autonomous engine

Architecture/engineering:

Design a scalable, maintainable architecture to replace the current monolith Define requirements for data, logging, observability, fault tolerance, and security Prototype critical components; run architecture and code reviews Establish technical standards and help decompose work into deliverable phases

Data/analytics/automation:

Design AI‑agent workflows for supplier requests, web search, and live price retrieval Define LLM/tool use, confidence scoring, and human‑in‑the‑loop controls and approvals Build exception handling (e.g., AOG cases, high deal value, low model confidence) Contribute to prompt logic, extraction pipelines, and agent orchestration Define transaction‑prioritization logic (margin, win probability, timing, customer/supplier risk, cash impact) Specify which decisions Copilot can make autonomously and which need human/C‑level approval Design next‑best‑action recommendations with auditability and explainability Pricing and decision‑making: Architect a Pricing Engine (margin, deal probability, supplier risk, commercial attractiveness) Separate rule‑based, statistical, and ML approaches Plan use of historical data (eMARS and other price sources) Select best price by recency, supplier reliability, lead time, and risk Management intelligence: Build a recommendation layer that explains what's happening, risks, options, and actions Define regular AI‑generated management reports (pipeline, margin, cash needs, customer/supplier risk, bottlenecks) Support SWOT analysis, budget allocation, and financing decisions Stakeholder management: Work closely with product, engineering, account managers, and senior management; translate business goals into workflows, architecture, and a phased plan Process and improvement: Maintain standards, documentation, mentoring, and continuous improvement Tech environment Core stack: Python (async), FastAPI, SQLAlchemy (or similar), DuckDB, SQLite (WAL), MySQL, PDF/XLSX/OCR parsing, LLMs (incl. Claude API), Tabulator.js

Infrastructure:

Event‑driven architecture, queues, asynchronous pipelines, CI/CD, logging/observability, security, fault tolerance

Integrations:

ERP/CRM, Gmail (service account), vendor databases (9,000+ suppliers), eMARS and other price sources, web search/retrieval; potential

ILS/PartsBase Collaboration:

Standard issue tracking, documentation, and communication tools Current system modules: ERP ingestion: MySQL adapter, RFQ input, daemon/manager Vendor selection: DuckDB matching, ranking, blacklist, 9,000+ supplier

DB Messaging:

Gmail service account, email generation, dispatcher, deduplication, cooldown Reply processing: Gmail auditor, PDF/XLSX/OCR parsing, Claude‑based extraction, confidence scoring

Storage:

SQLite WAL, sent registry, async audit logger

API/UI:

FastAPI, Tabulator.js,

Sent Emails Dashboard, RFQ Workbench Requirements:

Experience:

5+ years in backend development and architecture

Skills:

Python (async), FastAPI, SQLAlchemy or similar; event‑driven systems and queues; LLM APIs, AI agents, tool use; ERP/CRM integrations; evolving MVP/monoliths into scalable systems Architecture and integrations: End‑to‑end system design across RFQ→Close; agent/tool orchestration; human‑in‑the‑loop; auditability and explainability; external data/services integration Data and quality: Reliability, testing, security, observability; data/logging standards; confidence scoring and exception handling

Communication:

Engage business users and senior management; translate business needs into workflows, architecture, and a coherent roadmap; working proficiency in English As a plus: Domain knowledge: Procurement, B2B sales automation, supply chain; aviation MRO and aircraft components

Practices:

Recommendation/decision systems, human‑in‑the‑loop processes, operational analytics, cash‑flow forecasting, budget‑allocation tools, management recommendations Additional tools/certifications: Claude API; ILS/PartsBase; prompt engineering; web scraping and real‑time price retrieval; pricing/decision engines; account‑management automation; business‑process redesign What we offer: High impact on product and architecture; real ownership Flexible schedule; remote work A culture of experimentation, R D, and modern approaches Hiring process: Recruiter screening

• Technical/practical assessment
• Team and stakeholder interviews
• Final interview and offer