NEGAR (NINA) BAYBORDI

– Architected an end-to-end POC automation system that ingests meeting transcripts and produces structured documents through a multi-stage LLM pipeline (Amazon Bedrock, LangChain) featuring RAG retrieval (FAISS), prompt engineering, and a hybrid validation layer combining LLM-based semantic verification with cosine similarity scoring — delivering 90% reduction in documentation time and 98% formatting accuracy. – Built a conversational AI travel assistant on AWS Bedrock Agent to deliver personalized flight recommendations based on user preferences (budget, timing), processing 500+ queries with 85% accuracy. – Deployed with comprehensive error handling and validation workflows ensuring response accuracy and system reliability. – Built the FastAPI backend and Streamlit interface for a dental diagnostic AI application, integrating existing PyTorch/YOLO models for 5 dental conditions. Studied and adapted the model code to optimize preprocessing and confidence thresholds, enabling reliable model serving and interactive front-end visualization. – Deployed services using Docker and AWS container services (ECR, ECS), implementing CI/CD pipelines (GitHub Actions, ECR, ECS) and integrating serverless workflows with AWS Lambda, S3, IAM, CloudWatch, and Load Balancer to support scalable application operations. – Drove rapid experimentation and proof-of-concept (POC) initiatives, creating learn-fast frameworks to test hypotheses, validate AI/ML models, and identify high-impact opportunities aligned with business goals. – Provided AI thought leadership and mentorship, guiding new team members on LLM development, MLOps workflows, and data-driven strategy, ensuring alignment of AI initiatives with organizational objectives.

– Extracted and prepared data from web sources using Python (Selenium) for sentiment analysis, transforming unstructured text into analysable datasets. – Manually labelled a subset of the dataset to create ground truth for supervised learning, then used ALBERT embeddings to predict labels for the remaining data, ensuring high-quality labeled data for training. – Built sentiment classification models using ensemble methods (Random Forest, XGBoost) and deep learning (LSTM), achieving strong accuracy on imbalanced text data through systematic evaluation. – Improved data quality by identifying and correcting mislabelled samples, reducing model bias and strengthening class representation across sentiment categories. – Integrated processed data into AWS pipelines (EC2, S3, Glue, Athena) for automated reporting and designed Power BI dashboards and KPIs to monitor performance and provide actionable business insights.

– Improved delivery speed and transparency by managing Agile sprints in Jira, optimizing planning and resource allocation, and leading stakeholder meetings to align priorities and expectations. – Boosted product quality by conducting thorough QA testing, resolving major functionality gaps, and developing the first interactive software prototype to streamline user feedback and reduce development cycles. – Created dynamic reports using SQL, Power BI, and Dynamics 365 Reporting, helping teams make better decisions by improving performance tracking across departments and ensuring secure, role-based access to data.

– Administered Project Server and SharePoint, managing planning, reporting, permissions, and workflows while supporting users with system queries and troubleshooting for improved efficiency. – Analyzed project reports against Project Server (PWA) data to ensure accuracy and reliability, validating that data was sourced from the correct underlying tables and aligned with project performance metrics. – Led Project Server training sessions, promoting effective use of document management and collaboration tools, which enhanced team productivity and streamlined workflows. – Managed workflows and detailed project schedules using MS Project in line with PMBOK and PMI standards, integrating forecasting, budgeting, and risk analysis to deliver on time and within scope.