tehilla obanor

Translated launch and campaign requirements into automated AI content generation workflows. Built a Python orchestration layer standardizing requests and outputs across external AI providers, and implemented asynchronous job queues and asset management.

Built and maintained custom Python agents and n8n automations on a self managed VPS for sales, marketing, and operational workflows. Adapted workflows as requirements changed, using Git for version control and controlled releases.

Built a documented FastAPI NLP service that parsed CVs and extracted skills, fine-tuning spaCy NER to 85 percent F1. Used feedback from external testers to convert findings into monthly feature releases.

Deployed an isolated multi tenant RAG application with authenticated APIs and an embeddable JavaScript widget. Designed PostgreSQL and pgvector content isolation and implemented streamed responses and API health checks.

Worked directly with non technical stakeholders to clarify operational requirements and maintain a live reporting workflow for an event with over 60 attendees. Translated supplier and expenditure data into decisions, contributing to an estimated 30 percent reduction in projected costs.

Built an explainable financial risk machine learning pipeline using data from 68 countries covering 1980 to 2022. Compared attention LSTM, XGBoost, and Probit models, achieving 37 to 42 percent recall and 0.68 to 0.72 ROC AUC.

Built a Python, Pandas, and Docker ETL pipeline that cleaned and combined approximately 4,500 records across sales, logistics, production, pricing, and cost data. Developed dashboards that reduced manual reporting time by about 35 percent.

Reviewed operational and financial datasets to identify control failures, reporting inconsistencies, and potential fraud indicators. Prepared evidence-based reports against KPIs and control requirements.

Automated data extraction and consolidation using Python, SQL, and Excel, improving processing efficiency by 20 percent and reducing reporting turnaround time by approximately 15 percent. Developed forecasting models with more than 75 percent recorded accuracy.

MSc in Financial Technology from Coventry University, London, from 2024 to 2025.

BSc in Economics from Covenant University, Ota, Nigeria, from 2018 to 2022.