I'm Yavar Khan,
an AI Engineer.

AI/ML Systems

End-to-end machine learning pipelines — from custom Transformers and CNNs to anomaly detection autoencoders and reinforcement learning agents.

GenAI Applications

Multi-agent architectures, RAG pipelines, and LLM-powered assistants with structured outputs, tool use, and production-grade observability.

Scalable Software

Production REST APIs, cloud-native deployments, and monitoring dashboards — built for high-concurrency traffic and operational reliability.

Multi-agent async research system with query planning, multi-wave evidence collection, and cited synthesis across six specialized agents.

Intelligent RAG pipeline with FAISS vector indexing, runtime persona switching, and real-time Pushover alerts, deployed on Hugging Face.

Custom Transformer built from scratch in PyTorch with multi-head self-attention and Word2Vec embeddings, trained on 560K Yelp reviews.

Comprehensive comparison of VGG-16 and ResNet-18 on 30K images across 3 classes with hyperparameter tuning and data augmentation.

Convolutional Autoencoder for unsupervised anomaly detection in machine temperature data using the Numenta Anomaly Benchmark.

CNN for 36-class handwritten character recognition (26 letters + 10 digits) trained on 100K EMNIST balanced samples.

SARSA vs n-Step Double Q-Learning comparison on custom GridWorld with Optuna-based hyperparameter optimization.

1D CNN for classifying 5 emotions from speech audio using MFCC features, data augmentation, and a Flask web interface.