Looking for a job

Mark Kien User

@markkienuser

Message

I build production LLM, RAG, and fine-tuning systems that improve AI quality, cost, and scale.

Philippines

Message

I build production LLM applications, RAG workflows, AI agents, and model-training systems. At Ineffable Intelligence, I designed enterprise knowledge-discovery workflows spanning ingestion, retrieval, reranking, context construction, and grounded generation.

At Imobisoft, I built a distributed Transformer pretraining pipeline processing more than 1.2B training tokens per day and reduced unnecessary LLM calls by approximately 28%. I also fine-tuned open-source models with SFT, LoRA, QLoRA, and PEFT for domain-specific instruction-following tasks.

At Dynamo AI, I improved document-classification F1 from 82% to 93% through data augmentation, fine-tuning, and hyperparameter optimization. I bring hands-on experience across Python, PyTorch, Transformers, evaluation pipelines, GPU infrastructure, and production AI deployment.

Experience

Work history, roles, and key accomplishments

LLM / Data Engineer

Imobisoft

Jan 2022 - Dec 2024 (2 years 11 months)

Designed a distributed Transformer pretraining pipeline using PyTorch, FSDP, mixed precision, and multi-GPU training, processing over 1.2B training tokens per day. Fine-tuned open-source LLMs using SFT, LoRA, QLoRA, and PEFT for domain-specific tasks.

PyTorch Fsdp Mixed Precision Multi GPU Training SFT LoRAs QLoRA PEFT LLM Reinforcement Learning

AI Engineer

Dynamo AI

Jan 2021 - Dec 2022 (1 year 11 months)

Developed Transformer-based NLP systems for document classification, information extraction, semantic similarity, and question answering. Improved document classification F1 score from 82% to 93% through data augmentation, fine-tuning, and hyperparameter optimization.

Transformers NLP BERT Fine Tuning Document Classification Information Extraction Semantic Similarity Question Answering Benchmarking

AI / LLM Engineer

Ineffable Intelligence

Jan 2025 - Present (1 year 8 months)

Designed LLM-powered applications combining retrieval, embeddings, structured generation, and backend services. Architected RAG workflows for enterprise knowledge discovery and reduced fine-tuning GPU memory consumption by approximately 60% using QLoRA.

LLM RAG QLoRA Embeddings Structured Generation Backend Services Evaluation

Education

Degrees, certifications, and relevant coursework

Stanford University

Master of Science, Computer Science

2018 - 2020

Pursued a Master of Science in Computer Science with coursework in artificial intelligence, machine learning, deep learning, NLP, computer vision, algorithms, distributed systems, and databases.