Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m a Machine Learning engineer with 10 years of experience building production-grade NLP and LLM-driven systems. I’ve led the design and deployment of AI chatbots, RAG pipelines, and transformer-based models across enterprises, improving accuracy, latency, and user satisfaction. I stand out for bridging deep learning research with real-world applications, delivering reliable, scalable conversational AI solutions, and mentoring teams on best practices in LLMs and MLOps.…I’m a Machine Learning engineer with 10 years of experience building production-grade NLP and LLM-driven systems. I’ve led the design and deployment of AI chatbots, RAG pipelines, and transformer-based models across enterprises, improving accuracy, latency, and user satisfaction. I stand out for bridging deep learning research with real-world applications, delivering reliable, scalable conversational AI solutions, and mentoring teams on best practices in LLMs and MLOps.

Dan Thach

AI Engineer, Data Scientist, Developer, +2





I’m a Machine Learning engineer with 10 years of experience building production-grade NLP and LLM-driven systems. I’ve led the design and deployment of AI chatbots, RAG pipelines, and transformer-based models across enterprises, improving accuracy, latency, and user satisfaction. I stand out for bridging deep learning research with real-world applications, delivering reliable, scalable conversational AI solutions, and mentoring teams on best practices in LLMs and MLOps.…I’m a Machine Learning engineer with 10 years of experience building production-grade NLP and LLM-driven systems. I’ve led the design and deployment of AI chatbots, RAG pipelines, and transformer-based models across enterprises, improving accuracy, latency, and user satisfaction. I stand out for bridging deep learning research with real-world applications, delivering reliable, scalable conversational AI solutions, and mentoring teams on best practices in LLMs and MLOps.

Available to hire

I’m a Machine Learning engineer with 10 years of experience building production-grade NLP and LLM-driven systems. I’ve led the design and deployment of AI chatbots, RAG pipelines, and transformer-based models across enterprises, improving accuracy, latency, and user satisfaction. I stand out for bridging deep learning research with real-world applications, delivering reliable, scalable conversational AI solutions, and mentoring teams on best practices in LLMs and MLOps.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Work Experience

Senior AI/ML Engineer at Tkxel

April 1, 2024 - Present

Leading enterprise-scale LLM and Retrieval-Augmented Generation (RAG) system development; unified retrieval, generation, and observability to deliver auditable AI solutions. Designed and deployed an enterprise-grade RAG pipeline powering internal knowledge assistants, reducing document retrieval latency by 42% and hallucination rates by 35%. Standardized LLMOps practices across multiple teams, including model cataloging, retrieval evaluation, and lineage tracking. Built hybrid inference stacks combining open-weight LLMs (Llama-family) with hosted APIs for secure, cost-sensitive workloads. Established RAGOps pipelines for automated reindexing and freshness control across multimodal datasets. Collaborated with governance and security teams to implement PII-safe hybrid inference and prompt-based access controls.

Senior Machine Learning Engineer at Meta

March 1, 2021 - April 1, 2024

Drove development and deployment of LLM-based conversational systems, retrieval-augmented QA, and summarization models for enterprise applications. Built a production-grade RAG platform integrating OpenAI APIs with internal search, achieving a 28% improvement in response factuality and cutting inference costs by 22%. Led instruction-tuning and prompt optimization for in-house LLMs, improving user satisfaction scores by over 30% across QA and summarization applications. Designed hybrid cloud inference pipelines balancing latency and cost, leveraging caching, distillation, and quantized inference. Partnered with data operations to curate, label, and evaluate large corpora for domain-specific fine-tuning. Deployed model observability dashboards to monitor hallucination rates and retrieval coverage. Mentored a team of ML engineers on LLM integration, safety, and MLOps best practices.

Machine Learning Engineer at Semantic Visions

June 1, 2018 - February 1, 2021

Developed and productionized transformer-based NLP and CV models, advancing search and question-answering capabilities across large-scale platforms. Implemented BERT-based ranking and QA pipelines; automated fine-tuning workflows, data versioning, and A/B testing for continuous model iteration. Delivered compressed and distilled model variants to meet latency and throughput SLAs. Integrated sentence-transformer embeddings into semantic retrieval APIs powering downstream services. Collaborated on annotation and evaluation frameworks to maintain high-quality datasets for supervised fine-tuning.

Data Engineer at Featurespace

November 1, 2016 - May 1, 2018

Owned design and orchestration of large-scale data pipelines to accelerate ML experimentation and enable neural NLP adoption. Built automated ETL and feature pipelines using Airflow and Spark; maintained feature stores supporting early neural NLP experiments and recommendation systems. Developed embeddings-based NER and intent classification prototypes, replacing feature-heavy classical models. Partnered with ML teams to deploy early TF/Keras-based models through lightweight Flask services. Optimized streaming ingestion from Kafka to real-time analytics and feature layers.

Junior Data Engineer at UiPath

October 1, 2015 - November 1, 2016

Supported data infrastructure development and ETL pipelines during the early deep learning adoption phase. Built ETL pipelines consolidating clickstream and log data into Hadoop/S3 for downstream analytics. Developed prototype feature pipelines supporting basic supervised learning tasks such as churn and CTR prediction. Contributed to early deep learning proofs-of-concept models for image and text classification.