Jack of all trades, master of none
but sometimes better than a master of one
and yeah my initials are AI.
Analysed financial data across two client accounts in Excel, reconciling balance-sheet items and flagging discrepancies alongside senior consultants. First contact with audit methodology, risk procedures and how real client work gets delivered.
Building a nature-risk scoring platform for bank credit teams in Python, JavaScript and PLpgSQL. Mapped sectoral ecosystem dependencies at industrial-site level with NACE and ENCORE, and shipped a portfolio batch pipeline — SIRET input, INSEE SIRENE geocoding, WRI Aqueduct water-stress scoring, interactive Folium map out.
Driving organic growth for an AI startup across TikTok, Instagram, X and Reddit. ~1,000+ views/day (~7,000+/week) through a three-pillar content strategy — product demos, founder journey and scenario-based AI call clips — managing end-to-end production and performance iteration.
Replaced high-volatility static LTV tracking with a dynamic, state-conditioned framework. Built a Customer Migration Matrix from three years of transactional data across behavioural CM1 segmentation, transition-state discovery and LTV optimisation — and automated the weekly executive FP&A roundup.
An open-source CLI that turns Obsidian into an AI second brain. One npx command wires Claude into your vault: morning briefings written before you wake up, YouTube videos, podcasts and PDFs turned into structured notes, auto-generated profiles for people mentioned. Everything stays as plain Markdown on your machine.
A nature-risk scoring platform for bank credit teams. NACE and ENCORE databases mapped to sectoral ecosystem dependencies at industrial-site level, a batch pipeline from SIRET input through INSEE SIRENE geocoding to WRI Aqueduct water-stress scoring, and an automatic asset discovery layer built on INSEE corporate data, Mistral Large 3 and Exa Search.
