Tosin Kehinde

LLM Fine-Tuning & RLHF: Evaluate, rank, and rewrite AI-generated responses for multi-turn conversational agents, directly improving model truthfulness and reducing harmful hallucinations by 22%. STEM Content Validation: Author and verify complex math, coding, and physics training prompts, ensuring absolute factual accuracy and clear, step-by-step reasoning chains for specialized LLMs. Linguistic Quality Assurance: Serve as Lead Auditor for a team of 15 annotators, reviewing linguistic QA pipelines and maintaining a project-wide data accuracy rating of 99.1%. Chatbot Evaluation: Designed and executed adversarial testing (red-teaming) scenarios to identify safety vulnerabilities, successfully patching over 45 structural edge-case flaws prior to production deployment.

As a passionate and dedicated Customer Service Representative with 2 years of experience, I thrive on delivering exceptional service and creating positive customer experiences. My expertise lies in resolving complex issues, building strong client relationships, and consistently exceeding performance targets. I am committed to continuous improvement and always seek opportunities to enhance my skills and contribute to my team's success.

Front desk management Calendar management Correspondence management Document management Travel arrangement Office supplies management Data entry Expense reporting Basic administrative task

Search Engine & Query Relevance: Evaluated and scored search engine query responses for user intent and contextual relevance, boosting overall search algorithm accuracy scores by 14%. Computer Vision Annotation: Labeled over 50,000 image and video frames using bounding boxes, semantic segmentation, and keypoint tracking for autonomous vehicle and retail computer vision models. Social Media Content Evaluation: Annotate and categorize high-volume social media data for sentiment analysis, trend tracking, and content moderation, strictly adhering to complex, evolving policy guidelines. Audio Data Annotation: Tagged, segmented, and labeled multi-speaker audio files to train automatic speech recognition (ASR) models, focusing on dialect variations and background noise filtering.

Translation & Localization: Provided high-fidelity translation and cultural localization services for digital applications, ensuring semantic accuracy and preserving tone across English and Spanish variants. Audio/Video Transcription: Transcribed and time-stamped more than 800 hours of complex multi-dialect audio and video footage with an industry-leading 98.5% verbatim accuracy rate. Media Analysis & Sentiment Evaluation: Conducted deep-dive qualitative analysis on international marketing campaign media, labeling emotional triggers and intent to train automated brand-sentiment tools. Quality Benchmarking: Developed an internal transcription style guide that reduced onboarding time for new linguists by 30% and unified data formatting standards.