Oluwaseyi (Ekpere) Oyewunmi

Web scraping and anti-bot engineering: reverse-engineering how websites load and serve their data to get past advanced bot protection, including Cloudflare, Akamai and DataDome. Building and maintaining core product features: scraping, screenshot capture, data extraction and LLM content pipelines. Maintaining and scaling the backend infrastructure behind them to keep success rates high and latency stable across millions of requests against constantly changing targets.

Built an influencer analytics platform scraping 3K+ profiles daily across 5 social networks, reducing manual research time by 90%. Engineered engagement scoring algorithms processing 400K+ posts daily with sub-second query response times, powering data-driven influencer selection for 50+ active campaigns. Shipped NLP-powered query interface enabling non-technical team members to extract insights via natural language, reducing analyst bottleneck by 70%. Solo-architected full AWS deployment (EC2, Docker, CI/CD pipelines) from scratch, achieving 99.5% uptime with zero-downtime deployments.

Architected microservices backend (FastAPI + gRPC) improving API response times by 70% while maintaining 99.9% uptime through load balancing and fault tolerance. Migrated infrastructure to Docker + Kubernetes, reducing AWS costs by 80% and deployment time by 40%. Implemented Redis caching layer and database query optimization, reducing database load and improving system scalability. Mentored 3 junior engineers, establishing code review processes and testing standards that reduced production bugs.

Co-led the end-to-end build of BRaaS, a multi-tenant Telegram automation platform that guides moderators through enrollment, raid setup, help, and payout flows with guarded conversation handlers, role-aware routing, and graceful fallbacks. Engineered raid orchestration system coordinating Playwright-based Twitter scraping, AI-powered content scoring, and automated payouts via Firebase, processing 1K+ raids/week. Developed anti-detection Twitter scraper (Playwright) with cookie persistence and fingerprint rotation, achieving 95%+ success rate and exporting 5K+ engagement records daily. Built ML grading pipeline scoring grammar, sentiment, and marketing signals across 100K+ tweets, generating automated PDF reports for campaign stakeholders.

Developed scalable backend systems and ML pipelines using Python and FastAPI, optimizing transaction throughput and reducing model integration time across production environments. Architected cloud infrastructure achieving 99% improved uptime, implementing CI/CD automation with Docker/Kubernetes while deploying services on GCP (Google Cloud Platform) and managing VM environments on Proxmox. Championed a DevSecOps culture, integrating security practices into the CI/CD pipeline, which led to a 60% reduction in security vulnerabilities identified in prod environment.

Led development of Roulette Aid, an ML-powered desktop application predicting outcomes with 80%+ accuracy, resulting in 5x user engagement and significant revenue growth. Created ML models using scikit-learn and TensorFlow with continuous monitoring and retraining capabilities. Designed scalable backend architecture using FastAPI achieving 20ms average response time, with a fault-tolerant SQL database optimized for real-time data processing. Implemented advanced indexing, caching techniques, and multi-threaded processing that reduced computation time by 60%. Implemented comprehensive security through OAuth 2.0, code obfuscation, and encryption to protect proprietary algorithms while developing a tiered licensing system with secure FTP updates.

Architected a multi-tenant Django school platform, proxy personas for students/teachers/parents/head teachers, and relational profiles that tie teachers to classes/subjects and parents to wards, so every persona lands on a tailored dashboard experience, thereby revolutionizing academic workflows with enhanced student autonomy. Shipped e-learning module with class-specific digital library (500+ books), full-text search, and analytics dashboards tracking grades and engagement across 20+ metrics. Built ML model predicting student performance (89.8% accuracy) using scikit-learn, enabling early intervention for at-risk students across 1K+ enrollments.

Delivered 15+ client projects, including voting platforms, quiz systems, gym management tools, and automation solutions with 100% on-time delivery rate.