TAEYANG
KIM
Full-Stack AI Engineer with 4+ years building production ML systems from research to deployment. Specialized in rapid prototyping of multimodal GenAI applications with proven ability to scale (13% conversion lift, $2.4B portfolio). 9x hackathon winner and 5x US Patents holder bridging cutting-edge research (LLM, RLHF, Agentic Workflow) to production systems serving enterprise customers
MLOps • RAG • AgentCore • Robotics
Professional Impact
Senior Machine Learning Engineer at Pattern with 5 US Patents in Generative AI and Robotics. Specialist in architecting scalable, end-to-end ML systems. Recently showcased at AWS re:Invent 2025 for engineering an Agentic AI system that scales across 104,000+ SKUs. Proven ability to bridge model research with robust software engineering, delivering $16M+ in measurable business impact.
Experience Highlights
Senior Machine Learning Engineer
- Architected an agentic ReAct platform with CrewAI and Bedrock AgentCore showcased at AWS re:Invent 2025; built a fault-tolerant Redis + Airflow pipeline processing 104k+ SKUs, driving 13% conversion lift across a $2.4B portfolio
- Led a 10-engineer team to launch TrendVision (Multimodal GenAI Platform) from 0 to 1; architected MVP, shipped generative video/audio services on cloud with auto scaling for 200+ enterprise brands, achieving 6 min latency per video
- Engineered JAX-based RLHF pipelines using Group Relative Policy Optimization (GRPO) for automated prompt tuning, achieving an >80% preference rates from brand managers in blind A/B tests against baseline Amazon SEO copy
- Reduced manual workloads by 96% by building the Portal, a robotic capture system (Sony SDK, ggCNN, C++ server) that feeds real-time physical data into Gemini 3 model (Nano Banana Pro) and LoRA detailer workflows for image generation
- Improved IT support resolution time by 87% by implementing a hybrid RAG pipeline (semantic + lexical search) with Pinecone, OpenAI embedding model, Slack, Ruby, Snowflake and OpenAI GPT 4o used in production
- Built Flink + ClickHouse streaming pipeline processing 12M+ daily updates with continuous precomputation, reducing TikTok Shop latency by 95%; implemented auto-scaling compute and Redis caching for sub-second OLAP queries
Software Engineer
- Intelligent Agentic Orchestration: Developed an agentic tool-calling framework using MCP-like orchestrator and AWS Lambda, EventBridge; built event-driven action pipelines for real-time user interactions via Apple LiveChat and Twilio.
- Optimization: Boosted user engagement by 27% by architecting a scalable AI reminder API using AWS Lambda, SQS, and OpenAI.
- Scalability: Resolved 20+ critical bugs in message handling for 2M+ users across Apple LiveChat and Twilio integrations.
Machine Learning Engineer
- On-Device Model Research: Spearheaded R&D for 'insyt,' a 67M parameter distilled model (DistilBERT); optimized for high-performance intrusion detection on resource-constrained edge devices; achieved 89.7% accuracy.
- First Place Winner: Won 1st Place at the BYU Capstone Celebration Competition in Machine Learning.
- Architecture: Constructed an NLP framework using React, Keras, Redis, and Flask to classify network intrusion; performed N-Gram EDA on AIT Log Data Set V2.0.
- Deployment: Optimized and deployed the library on HuggingFace and PyPI for global access.
Software Engineer
- Full-Stack Development: Built administrator dashboards and premium test result pages using GraphQL, Google Firebase, Firestore, Next.js, and TailwindCSS.
- Data Analytics: Integrated Split.io and Google Analytics for advanced user behavior tracking and ad tracking.
Software Engineer
- Spearheaded a deployment of a real-time audio editor software for seamless text-to-speech and speech-to-text conversion using Amazon Transcribe, AWS Lambda, Python, React.js, and Node.js.
Research Assistant II
- Small Language Models (SLM): Developed a full-stack application utilizing Transformer-based SLMs (GPT architecture) using PyTorch, Neo4J, React, and Flask.
- Computer Vision & Robotics: Invented an autonomous farming robot featuring Lidar, humidity sensors, and real-time vision-based weed/fertilizer management.
Software Engineer
- Created a real-time business dashboard displaying critical ad hoc data with an authentication, REST APIs, and UI through React.js, Next.js, TypeScript, Rust, PostgreSQL, Supabase, and TailwindCSS.
- Expertly organized a relational database and RESTful APIs for CRM applying Rust and PostgreSQL.
Levinthal Fellow | Software Engineer
- Selected as one of 15 Levinthal Fellows ($2000 Scholarship) for high-impact tech startups in the Sandbox incubator program.
- Developed and competed at the BYU Sandbox hackathon, focusing on entrepreneurial stories and industry insights.
Software Engineer
- Infrastructure as Code (IaC): Built a serverless pipeline for Video/Audio Stream Save & Conversion using AWS CDK 2, ECS Fargate, Lambda, and Step Functions.
- CI/CD Optimization: Reduced deployment latency by 40% (40 min to 24 min) through refactored CDK pipelines and Docker-based builds.
- Testing & QA: Executed comprehensive Unit, Integration, and E2E test suites using Amazon CloudWatch, Jest, and Axios.
- Live Video Innovation: Implemented a Live Video Stream Rewind feature (YouTube-like) using EdgeCast API, Video.js, and Vue.js.
- Mobile Development: Contributed to the release of BYU TV and BYU Radio Android applications using Kotlin and RxJava.
Operational Specialist
- Operational Leadership: Assisted operation orders at the Eighth Army NCO Academy; evaluated training performance for 1,800+ soldiers annually in BLC (Basic Leadership Course) and KATUSA programs.
- Technical Documentation: Defined and wrote comprehensive manuals for unit tasks and individual soldier responsibilities, managing 18-20 soldiers in Gyeonggi, South Korea.
- Awards: Received Army Commendation Medal (ARCOM) from Eighth Army HQ BG and LTG Bills, General Paid Sun Yup Leadership Award, and KATUSA of the YEAR 2019.
Featured Projects
🏆 1st Place Winner – Utah AI Hackathon. AI-powered therapy robot providing emotional support through real-time speech recognition.
🏆 1st Place – Utah Developer Hackathon. AI-driven roofing analysis and target marketing platform using Google Maps API and Vision Models.
🏆 1st Place – BYU Capstone Celebration. LLM-powered network intrusion detection system for Sandia National Laboratories.
🏆 1st Place + Benny Award – BYU ACM YHack. Automated viral shorts generation from long-form video using microservices.
Google Gemini Hackathon entry. Transforms product photos into compelling marketing videos using Gemini 1.5 Flash.
AI-powered matching platform used by 2025 YSA Conference participants. Published on Apple App Store & Google Play.
Let’s Engineer
Something Great.
Currently open to AI Software Engineering roles and Research collaborations.
k2289@byu.edu
Phone
+1 385 283 7793
Location
Provo, UT, USA