Raphael Kim

AI Systems EngineerConcept to buildSoftware + hardware

AI Systems Engineer. From concept to working system.

My educational background spans mechanical engineering at ENS, virtual reality engineering at ENSAM, and industrial design at Strate.

I conceive, design and build AI-powered systems — connecting models, tools, software and hardware, from the first idea to a working product.

Voice systems at Snips / Sonos. Collaboration prototypes at SAP. Physical interaction research at MIT Media Lab. Today, I bring that experience to local AI, agentic workflows and connected objects.

Selected work

Systems, objects and experiments I conceived and built. Try the interactive studies, then open each project for the full story.

What I bring to a team

I work across the boundaries where a promising demo becomes a usable system: interaction, orchestration, inference, persistent state, embedded hardware and deployment. I can define an experience, make the technical choices and build the pieces that let someone actually use it.

The projects here originate in my own concepts and design work. I take them through implementation and iteration; professional contexts and team contributions are identified where relevant.

Built in different contexts

  • Snips / Sonos — voice and product prototyping. Contributed to the product definition and UX specification of Sonos Voice Control. Conceived and built contextual listening prototypes spanning interaction, lyrics retrieval, sensing and ESP32 firmware.
  • SAP — collaboration. Conceived the idea, built the prototype and developed the OS X application within the Chairman’s project team. The concept later evolved into CoScreen.
  • MIT Media Lab — physical interaction. Created SPrAyCE, a spray-based interface for drawing in mid-air, connecting a physical gesture to digital creation.
  • Independent R&D — AI systems. Design, build and operate the local inference, voice services, agent integrations and interfaces behind the projects in this portfolio.

How I build

Start with a use case. Make the whole loop work. Then find the weak boundaries: ambiguous requests, missing context, unavailable services, unsafe actions and outputs that cannot be verified.

I use models where interpretation or generation helps, and explicit software contracts where correctness matters. That can mean a source-backed answer, a reviewed diff, an event ledger or a hardware control loop that keeps working without the model.

Synapse architecture

Local-first AI, from an interaction to durable work. Specialized models and independently operated services compose the system.

Interfaces

  • Telegram
  • Web
  • Voice
  • ESP32

Capabilities and permissions vary by interface.

Decision & orchestration

  • Interaction Decision
  • Primary
  • Task / DSH

Reply, call a service directly, or delegate durable work.

Intelligence & capabilities

  • LLM Fabric
  • Speech
  • Vision
  • Content / Knowledge
  • Tools

Specialized services

  • Home Assistant
  • Media / Reading
  • Archivist
  • Context Observer
  • Observability
  • Watch

LOCAL COMPUTE

Fabric · model routing & inference access

01 / TEXTDEPLOYED

ia01

ASUS GX10 · DGX Spark-class
NVIDIA GB10 · unified memory

MODEL / SERVING

Qwen 3.8 Flash Next

vLLM · single-node inference

Primary reasoning & tool use
02 / VISIONDEPLOYED

ia02

ASUS GX10 · DGX Spark-class
NVIDIA GB10 · unified memory

MODEL / SERVING

Nemotron 3.5 Super VL

vLLM · dedicated image inference

Fabric availability unresolved at verification
03 / SPEECHDEPLOYED

ms03

MS-02 Ultra
NVIDIA RTX PRO 4000 Blackwell SFF

MODELS / SPEECH PLANE

Parakeet · Whisper
Kyutai · TitaNet

NeMo · CTranslate2 · Moshi

ASR · TTS · diarization · realtime voice
04 / ON DEMANDOPTIONAL

Workstation

NVIDIA RTX 5090 · 32 GB
Specialized compute & experiments

CONFIGURED MODEL / SERVING

Qwen 3.8 · 27B

NInfer · on-demand provider

Offline at verification · available when online
Compose capabilities. Keep execution durable. Primary can call specialized services directly or delegate to Task / DSH. Persistent state supports recovery; services keep their own execution boundaries.
LOCAL MODELS / MODULAR SERVICESDEPLOYMENT SNAPSHOT · OCT 2026