Tzechung K.

Work experience

• Artificial Intelligence Machine Learning Engineer, Product Manager, ResearcherOct 2024 – Present2 y.

  Developed end-to-end AI-powered ad engine leveraging LLMs and AI agents to optimize lead generation workflows and ad serving performance

  Responsibilities:

  • Created GenAI applications for synthetic data generation (text-to-image, video, captions) from multiple LLMs to support ad content creation and model fine-tuning
  • Built production-ready AI agents using LangChain, LangGraph, n8n, and a company Agent Development Kit for automated content generation and lead qualification. Also with MCP server
  • Implemented RAG systems with vector databases to enhance ad targeting and personalization, reducing latency by optimizing data streaming pipelines
  • Developed LLM evaluation frameworks using DeepEval, Claude Eval, and a company Vertex AI Eval to measure model performance and bias-variance tradeoffs
  • Fine-tuned custom LLM models using LoRA for domain-specific ad content generation and lead scoring applications
  • Researched and implemented On-Device AI solutions using TensorFlow Lite and ExecuTorch for edge deployment, enabling real-time ad serving on mobile devices
  • Utilized GCP stack including a company Colab, Kubernetes Engine, Cloud Run, and BigQuery for scalable ML infrastructure
  • Conducted computer vision research with AirFlow and Kafka for processing real-time device streaming data to enhance ad relevance
  • Creating a New Ad Engine derived from the one I created 15 years ago to include GenAI with fine-tuning LLM, RAG, AI Agent, supervised, unsupervised learning, postgreSQL and vector database, web (React) and Node.js
  • Using Azure Databricks, Snowflake, Delta Lake and Spark SQL create datasets for AI/ML modeling.
  • Include AI Safety guardrails to models and agents to detect prompt injection, prevent jailbreak

  Technologies: PyTorch, TensorFlow, LangChain, LangGraph, n8n, OpenAI API, a company Llama, a company Gemini, Anthropic Claude, GCP, Databricks, PySpark, React, Node.js, postgreSQL

• Machine Learning EngineerJun 2019 – Sep 20245 y.

  Scaled ML platform from inception to 1,500+ active users, supporting a company Firefly's generative AI development on a company A100/V100/H100 GPUs

  Responsibilities:

  • Collaborated cross-functionally with product managers, researchers, and data scientists to architect and deploy scalable ML infrastructure on AWS with RunAI orchestration
  • Developed model serving APIs using FastAPI for inference endpoints, enabling seamless integration with production applications and A/B testing frameworks
  • Implemented distributed training pipelines using data-parallel and model-parallel computing (FSDP/DDP) for large-scale diffusion models and LLMs
  • Built automated QA systems for ML platform validation, ensuring 99.9% uptime for critical training workloads
  • Created AI model evaluation framework as part of evaluation team, developing code to assess bias-variance, performance metrics, and comparative model analysis
  • Optimized GPU utilization through Kubernetes cluster management and a company-smi monitoring, reducing training costs by 30%
  • Partnered with On-Device team to deploy models to edge devices using TensorFlow Lite and model quantization techniques
  • Processed large-scale datasets using Databricks and DataLoader, implementing efficient data ingestion pipelines for continuous model improvement
  • Developed web-based model demonstrations using Flask and Docker containerization for stakeholder presentations
  • Provisioned datasets from a company asset using Snowflake, SQLite, PostgreSQL and AWS S3
  • Key Achievement: Core contributor to a company Firefly launch, one of a company's flagship GenAI products serving millions of users

  Technologies: PyTorch, CUDA, Kubernetes, RunAI, AWS, Databricks, FastAPI, Flask, Docker, a company H100/A100/V100

• Machine Learning EngineerNov 2018 – Jun 20197 m.

  Provided technical training and onboarding support for a company's ML platform to researchers and engineers

  Responsibilities:

  • Performed quality assurance and platform validation to ensure reliability for production workloads
  • Created documentation and best practices guides for ML workflow optimization
• Big Data EngineerApr 2018 – Oct 20186 m.

  Migrated 65+ million credit card records from Teradata/Snowflake SAS programs to AWS S3 using Apache Spark and Python

  Responsibilities:

  • Built data pipelines and CI/CD workflows on Databricks for automated integration testing and deployment
  • Optimized Spark jobs for performance, reducing data processing time by 40% through efficient partitioning strategies
  • Developed ETL processes using PySpark, SQL, and MLLib for real-time analytics on customer behavior data

  Technologies: Apache Spark, Databricks, Python, SQL, MLLib, Jenkins, GitHub, AWS S3

• Researcher, Data Scientist, Product Manager, Machine lerning Engineer2004 – Mar 201814 y.

  Built marketing analytics models for pricing optimization and ad performance monitoring using statistical analysis and A/B testing

  Responsibilities:

  • Created full-stack web applications including e-commerce platforms and classified ad websites with integrated ad engines
  • Established and managed on-premises data center with complete infrastructure stack for rapid prototyping and development

  Technologies: Python, Apache Spark, Azure ML, Databricks, SQL, Tableau, ASP.NET

• Senior Product Manager, Software Engineer II2004 – Jan 201713 y.

  Led full software development lifecycle for air defense systems serving international customers, from requirements gathering through deployment

  Responsibilities:

  • Managed cross-functional teams of software engineers, hardware engineers, and operations to deliver complex defense systems
  • Designed and deployed customer data centers with complete network, server, and application infrastructure
  • Conducted regression testing, integration testing, and UAT with automated test frameworks for production deployments
  • Developed proof-of-concept systems based on customer requirements, translating business needs into technical specifications
  • Security Clearance: Top Secret (held until 2000)
• Product Manager, Founder, Software Engineer2002 – 20042 y.

  Founded startup developing e-commerce platform with integrated ad engine and live-streaming capabilities (iCast.com) using VB.NET, a company SQL Server, and a company Video Player

• IT Manager2000 – 20022 y.

  Managed team of 10 engineers developing customer support systems and web applications using ASP.NET and MS SQL Server

• Marketing IT Manager1997 – 20003 y.

  Led marketing IT functions including sales forecasting systems and marketing websites. Received Best IT Support Manager achievement award from General Manager

• Founder1994 – 19973 y.

  Established startup generating $3.5M annual revenue through innovative ID card systems (IDMaker). Led cross-functional team developing mortgage fraud detection database in a company

• Manager - Financial Systems1978 – 19879 y.

  Managed team of 30 engineers developing enterprise billing system for 1,500 users. Delivered projects within budget ($45M) and on schedule across 3-year development cycle