AI / ML Engineer

Designed LLM-powered applications integrating retrieval, embeddings, structured generation, and backend services. Architected enterprise RAG workflows covering document ingestion, chunking, retrieval, reranking, context construction, and grounded generation.

Reduced fine-tuning GPU memory consumption by approximately 60% through QLoRA, enabling larger models to train on available infrastructure. Developed evaluation workflows for answer quality, retrieval relevance, latency, reliability, and cost, and built production multi-agent workflows using MCP and Hermes Agent with API/database integrations, memory, planning, error recovery, and monitoring.

Designed a distributed Transformer pretraining pipeline using PyTorch, FSDP, mixed precision, and multi-GPU training. The pipeline processed more than 1.2 billion training tokens per day.

Fine-tuned open-source LLMs using SFT, LoRA, QLoRA, and PEFT for domain-specific instruction-following tasks. Also reduced unnecessary LLM calls by approximately 28% for high-volume agent workloads and experimented with reward-model-based training and reinforcement-learning workflows.

Developed Transformer-based NLP systems for document classification, information extraction, semantic similarity, and question answering. Fine-tuned BERT and other Transformer models on domain-specific datasets.

Built automated benchmark pipelines to evaluate instruction following, reasoning, extraction, summarization, and other domain-specific tasks. Improved document-classification F1 score from 82% to 93% through data augmentation, fine-tuning, and hyperparameter optimization.

Relevant coursework included Artificial Intelligence, Machine Learning, Deep Learning, Natural Language Processing, Computer Vision, Algorithms, Distributed Systems, and Databases.