Daniel Chae

Building voice AI

Kubrick is an open-source deployable ingestion pipeline and API for multimodal semantic search over video, enabling users or RAG systems to retrieve video segments using natural language, images, audio, and video queries. - Prototyped and created an automated data pipeline that leverages event-driven architecture to scalably and reliably generate vector embeddings for video content (S3, SQS, Lambda). - Designed and implemented an auto-scaling RESTful API allowing clients to easily call video ingestion and search endpoints (AWS API Gateway + Lambda). - Developed a guided CLI tool enabling easy automated infrastructure deployment to AWS (Terraform, TypeScript). - Reduced search latency by 60% by refactoring S3 API calls to use asynchronous threads and optimising database queries with an HNSW aNN index (Python, Lambda, PostgreSQL, pgvector). - Built an interactive “playground” web app allowing users to experiment with the Kubrick API (S3 Static Sites, CloudFront, React, Next.js). - Engineered an observability layer to expose real-time insights into video processing and vector embedding workflows (Lambda, SQS, and PostgreSQL). - Collaborated with a team of four engineers across three countries, running daily standups and coordinating agile feature development.

Developed freelance and open-source applications including: • rBasket - Self-hosted tool for collecting and debugging webhooks. Built with DO Droplets, Nginx, MongoDB, Express, and React • Canter - A modular Neovim code runner plugin integrating Snacks terminal (Lua). • clementine - A terminal-based AI coding assistant and pair programmer built with Mastra, Ink, and React.

- Served as the main point of client-facing contact, successfully managing the entire product lifecycle across teams to ensure timely delivery of final image assets for multiple high-profile Netflix titles. - Developed and led trainings in technical digital asset management workflows (PhotoMechanic and DFStudio). - Provided live asset management and editing services, supporting the Netflix marketing team for live events and premieres.

- Was responsible for ingesting, standardising, culling, and managing the storage of still image assets including unit, gallery, BTS, and press photography for various film and series productions.

- Operation and maintenance of both computer and manual firing systems and pneumatic loading systems for the K9 self-propelled howitzer. - Short-term deployment to Incheon Airport Quarantine Office as an AQSF translator, working with KDCA to minimize introduction and transmission of COVID-19 from foreign countries. Honors: Division Commander Commendation: 2020/03, 2020/06, 2021/05, ‘Special Force’ Denomination 01/2021

- Portrait, fashion, wedding, event photography. - Developed photoshoot concepts and coordinated schedules across creative teams (stylists, makeup artists, models) to ensure on-time project delivery. - Collaborated with clients throughout the creative process to deliver polished, finished images.

Wrote custom code to tie coaching feedback to hundreds of character/situation-specific triggers for the development of a visual AI engine designed to automate competitive gaming coaching.

Introduction to Programming MOOC (Python) - Full Marks Advanced Course in Programming MOOC (Python) - Full Marks