Abdoulaye Saliou

Engineered complex system prompts and conditional logic to govern and test advanced agent behavior. Audited API parameters and JSON structures to diagnose flaky AI-generated tool calls and transformed model failures into high-quality RLHF training data. Conducted rigorous QA by cross-validating the agent's natural language summaries against raw JSON outputs to identify and rectify factual hallucinations, data omissions, and logical inconsistencies. Performed root-cause analysis on AI model failures across parameter errors and procedural missteps in multi-step reasoning tasks. Authored expert, structured feedback (Critic & Reasoning Comments) to create superior training examples for RLHF pipelines. Designed strategic prompts to stress-test model capabilities and generate targeted training data for areas like Search Refinement and DateTime Reasoning.

Spearheaded end-to-end development of mobile and web applications; mentored two junior developers on software development best practices and Agile methodologies; optimized UI to reduce navigation time by 25%; implemented a Firebase push notification system and integrated a secure payment gateway; architected and deployed a Node.js/Express-based microservices backend on AWS, reducing latency; automated CI/CD pipelines with GitLab.

Led sprint planning, task prioritization, and unblocking of development efforts. Developed responsive UX/UI mockups with Adobe XD. Built robust web services for banking and mobile money applications, focusing on rapid problem resolution and high client satisfaction. Optimized backend performance and security by refining complex SQL queries and improving overall system reliability.