• Finetine machine translation model on low-resource language with GAN and unsupervised learning.• Train a grounded open QA Chinese large language model from different architecture including T5, Llama, etc.• Optimize VRAM usage and training speed with ZeRO and quantization technique under multiple GPU setup• Preprocess and analyse data crawled from various web content to finetune our own Chinese LLM•

• Compiled different model system from source for scientific simulation software including wrf-hydro, deepMD, etc.• Optimized scientific simulation program and Run with different workload on HPC multi-node cluster• Reduced VRAM usage of Large Chinese NLP model, Yuan, by flash attention and 8-bit quantization•

• Implemented OAuth2 support for account from Shopify, Google, etc. to increase UX of our low code solution• Improved security of internal APIs by adding authentication checking and validation request payload•

• Developed various micro-services including GraphQL, Webhook for content distribution and chat function• Designed and deployed architecture of microservice using AWS S3/SNS/SQS/ECS/Route 53/Cloudfront• Configured and implemented Github Action/Dockerfile/Makefile for CI/CD and Github repo template for start-off