Adrianh Dao Senior Machine Learning Engineer

P

Senior AI/ML Consultant

Prudentia Sciences

Mar 2025 - Current (1 year 6 months)

Evaluated multiple vector databases including PGVector, OpenSearch Serverless Collections, and Pinecone to optimize indexing and querying performance. Achieved 150ms query time and 200ms indexing time (for 32 documents) with the best cost-efficiency using a serverless architecture.Developed a benchmarking framework to iteratively test and tune vector database configurations (e.g., HNSW parameters) for optimal performance across use cases.Diagnosed performance bottlenecks in the data pipeline and vector database integration, proposing scalable and cost-effective solutions aligned with the startup environment.Designed and implemented a microservices-based architecture to manage complex data ingestion and retrieval-augmented generation (RAG) p

Staff Machine Learning Engineer

eSimplicity

Nov 2023 - Mar 2025 (1 year 5 months)

Fine-tuned LLM (LLAMA 3.1 8B) to convert natural language queries into Cypher queries, enabling a GraphRAG pipeline over AWS Neptune. Achieved 63.32% Google BLEU and 31.57% Exact Match accuracy—surpassing ChatGPT-40 performance on internal benchmarks.Utilized Unsloth for parameter-efficient fine-tuning (PEFT) with Hugging Face Transformers, reducing GPU memory usage by 70% and cutting training time in half on AWS SageMaker.Designed and orchestrated serverless ingestion and retrieval pipelines using AWS Step Functions and Python Lambda, integrating heterogeneous sources including Confluence, PDFs, Excel, and Centralized Data Repositories (CDR) into a unified knowledge graph.Significantly improved ingestion quality by integrating Mineru to co

Senior Machine Learning Engineer

iFIT

Apr 2021 - Nov 2023 (2 years 8 months)

Spearheaded the deployment of a foundational LLM using Amazon SageMaker, fine-tuning LLMs to power NASM and ASCM certification programs, as well as enhancing the iFIT AI Coach chatbot.Optimized AI-driven recommendations by fine-tuning Amazon Personalize, significantly improving user personalization and delivering precise workout suggestions.Developed a highly scalable, real-time vector search engine using MongoDB Atlas to enable lightning-fast, contextually accurate recommendations for workouts, fitness plans, and health insights.Engineered intelligent text classification and semantic search algorithms, dramatically improving the precision of AI responses and optimizing search retrieval speeds by over 300%.Implemented a multi-tiered data st

Machine Learning Engineer

Apple

Jul 2019 - Apr 2021 (1 year 10 months)

Designed and developed an on-device question-answering application powered by a fine-tuned BERT transformer model, enabling real-time answers from long-form text without internet access.Trained and optimized BERT using TensorFlow and exported to CoreML format using TFCoreML, leveraging Core ML 3's transformer support for seamless deployment on iOS.Implemented custom WordPiece tokenization pipeline using Apple's NaturalLanguage framework to match BERT's input requirements and align token indices with original text spans.Built a full NLP pipeline utilizing speech framework for voice-to-text question input, Core ML for on-device model inference, and AVFoundation for text-to-speech answer playback, enabling a fully interactive QA experience.Opt

Machine Learning Engineer

DataRobot

Jul 2016 - Jul 2019 (3 years 1 month)

Developed and leveraged innovative features and algorithms to drive down false positives and identify perceived threat across the claims.Utilized traditional statistical analytics, graph theory / network science, ensemble methods and Natural language processing, text analytics, factors analysis / construct development and testing, machine learning feature development and engineering, etc.Employed ensembled methods to increase the accuracy of training model with different Bagging and Boosting method (Xgboost, LGBM, Gradient Boosting, Adaboost).Used cross-validation to test the models with different batches of data to optimize the model and prevent overfitting.Worked on feature selection and feature engineering to determine features that have

Machine Learning Intern

DataRobot

May 2015 - Jul 2016 (1 year 3 months)

Owned the whole life cycle of a machine learning model in production, from data collection and analysis, feature extraction, model training, evaluation, deployment, and refresh and maintenance.Trained the developed models and run evaluation experiments on customers reviews.Performed statistical analysis of results and fine-tuning models.Professional competency in Statistical NLP / Machine Learning, especially Supervised Learning - Document classification, information extraction, and named entity recognition in-context.Worked with Proof of Concepts (POC's) and gap analysis and gathered necessary data for analysis from different sources, prepared data for data exploration using data wrangling.Strong SQL Server and Python programming skills wi

Adrianh Dao

About

Detail

Timeline

Résumé