Rayane Oubelkas

Krappa

Founder & Applied AI Engineer · Remote · July 2025 - Present

• Built an Excel AI add-in with Office.js tools to create, audit, format, chart, and repair financial models; the system outperformed ChatGPT for Excel and Claude for Excel in internal agent evaluations.
• Built a PowerPoint AI add-in that converts HTML/CSS into editable finance and equity-research decks with native charts, source-backed templates, and automated slide QA; shipped 12 validated templates spanning 143 slides.
• Built Grinder, an Electron-based agentic finance workspace with primary-agent delegation to specialized subagents and secure access through Gmail, Outlook, WhatsApp Business, and Telegram.
• Integrated session-scoped Excel, PowerPoint, and PDF tools to create, edit, inspect, and validate financial models and presentations, with artifact tracking and safe file boundaries.
• Built a user-agentic plugin platform that lets users develop and install custom plugins bundling prompts, agents, skills, tools, workflows, hooks, context providers, scoped memory, and UI metadata.
• Built an internal agent-evaluation harness with 510 cases spanning DCF, LBO, three-statement, debt-schedule, and other financial-modeling tasks; measured pass rates, formula accuracy, formatting quality, tool-call count, token usage, latency, retries, and estimated API cost.
• Designed a leakage-resistant Qwen3.5-9B distillation pipeline combining verified tool SFT, skill-guided ReST, and preference/recovery training data for diverse financial-modeling tasks across industries, layouts, and model families.
• Generated and audited 35 industry/layout specialist skills and deterministic task/oracle families; enforced hash-bound checks for formulas, accounting integrity, artifact preservation, visual quality, screenshots, and provenance.
• Implemented 417 deterministic regression checks for accounting, formulas, reconciliation, spreadsheet integrity, and professional layout, preserving reproducible run traces.

Internal counts and comparisons are self-reported.

Locus Capital

Trader Intern · Remote · March 2025 - September 2025

• Built Python backtesting frameworks and applied quantitative market research and financial models to identify and validate investment opportunities.
• Built an AI market-research agent scheduled to run every three hours, retrieve market news through yfinance and financial data from the Financial Modeling Prep (FMP) API, and produce recurring research reports.

Tratop

FP&A Intern · Agadir, Morocco · June 2024 - August 2024

• Prepared budgets and financial reports for a topographic studies office using financial modeling and data analysis.
• Built structured budget models to organize operating assumptions and project-level financial inputs.
• Analyzed financial data and converted findings into clear reporting summaries for internal review.

Softron Tax

Tax Preparer · January 2024 - May 2024

• Prepared and filed simple personal income tax returns for individuals and families with modest incomes; provided clear, respectful, and community-centered guidance throughout the tax filing process; and ensured accuracy, compliance, and strict confidentiality when handling sensitive financial and personal information.