Irfan Aariz

Architected and deployed multi-agent Generative AI solutions using LangChain and vector databases (Pinecone, FAISS) to fully automate complex document intelligence and data query pipelines for Fortune 500 enterprise clients. Led the full lifecycle development of LLM fine-tuning pipelines using GPT-Neo and Mistral, resulting in reduced model hallucinations and increased conversational accuracy for key product features. Directed a 6-person ML Engineering team to deploy scalable, secure AI workflows across multiple cloud regions (AWS, GCP). Spearheaded MLOps transformation by implementing CI/CD automation with Docker, Airflow, and MLflow. Designed and implemented a high-performance Retrieval-Augmented Generation (RAG) framework utilizing hybrid search techniques with AWS OpenSearch and Redis VectorDB for low-latency knowledge retrieval.

Developed and operationalized advanced NLP systems (text summarization, entity extraction, sentiment analysis) that enhanced knowledge management platforms used by over 50 enterprise clients. Built and deployed high-throughput FastAPI-based inference APIs capable of serving 100,000+ daily requests with a guaranteed latency under 50ms. Implemented TensorFlow-based deep learning models for real-time anomaly detection in high-volume enterprise IoT data streams, improving uptime. Automated model retraining and monitoring pipelines to ensure continuous deployment and adaptive learning.

Assisted in the development and validation of predictive models for customer churn and marketing ROI optimization, directly influencing multi-million dollar campaigns. Conducted rigorous EDA and feature engineering on large-scale retail datasets using Pandas, NumPy, and SQL. Developed and maintained interactive executive dashboards using Plotly and Dash to communicate complex data insights to senior leadership.