Projects

AI Engineer focused on biomedical ML, LLM systems, and scalable training — targeting roles at Big Tech and Pharma.

Biomedical & Computational Pathology AI

  • Multimodal alignment for 7 modalities under missing-modality settings
  • Heterogeneous fusion for survival prediction (MIDL 2026)
  • Protein foundation model fine-tuning and structured knowledge injection
  • WSI preprocessing, patch sampling, and multimodal fusion pipelines

LLM & RAG System Engineering

  • Hierarchical RAG with multi-level document abstraction (8,000+ PDFs)
  • Document ingestion: Grobid, LlamaIndex, Linq-Embed-Mistral, Vicuna-13B
  • MapReduce summarization and semantic segmentation workflows
  • RAGAS evaluation framework — 25.73% improvement over baseline

Large-scale Training & Optimization

  • Distributed training (DDP), mixed precision (AMP), FlashAttention
  • Multi-task learning and gradient conflict analysis
  • Experiment tracking and custom training loops

Model Efficiency & Recommender Systems

  • Low-rank feature interaction layers (FiBiNet++, CIKM 2023)
  • 12–16x model size reduction with 37.5–81% efficiency gains
  • Benchmarking: DNN, DeepFM, xDeepFM, DCN, AutoInt, FiBiNet

AI Engineering Skills

  • ·PyTorch, TensorFlow, HuggingFace Transformers
  • ·Fine-tuning (LoRA/PEFT) & multimodal systems
  • ·RAG systems, embedding models
  • ·Distributed training (DDP), mixed precision (AMP)
  • ·Large-scale data preprocessing
  • ·Docker, AWS, GPU training, experiment tracking
  • ·feature engineering
  • ·evaluation frameworks