Hi, I'm looking for an experienced AI infrastructure specialist — or a passionate enthusiast with solid hands-on experience — to build a robust, flexible, and high-performance backend foundation for OpenClaw (an open-source autonomous AI agent) on two NVIDIA DGX Spark systems connected via 200 GbE (ConnectX-7). The goal is a stable, demo-ready environment that showcases the power of open-source models. OpenClaw will run on a separate system and connect easily via OpenAI-compatible APIs (or equivalent best-practice interfaces). Everything should prioritize:
Easy remote access and external connectivity (via Tailscale/ZeroTier). Fast performance within the hardware's unified memory constraints. Simple model switching/adding later (hot-swappable where possible). Persistent services with web UIs for live demos. On-demand tools that can spin up/down cleanly.
Hardware & Current State
2× NVIDIA DGX Spark (Grace Blackwell, 128 GB unified LPDDR5x memory each, ARM64). 200 GbE interconnect + 1 GbE internet links. I have a basic cluster setup, timeshift snapshots, ZeroTier, and Tailscale already running. You're welcome to rebuild from scratch if that's cleaner and faster.
Core Requirements (Persistent Where Possible)
vLLM as primary inference engine with a large-context main model (e.g., Nemotron 120B or equivalent). Must support easy switching to newer models. Whisper (or best-practice alternative like faster-whisper) – ready for OpenClaw API integration. Piper TTS – ready for OpenClaw API/text-to-voice integration.
All persistent services should run with clean web UIs for demo purposes. On-Demand Tools (Configured for Easy External/Tailscale Access + Web UIs)
Ollama + web UI (for specific or scheduled models). OCR model + workflow (let's discuss the best option—e.g., EasyOCR/PaddleOCR—and data saving/integration with other tools). Image generation (primary for OpenClaw use) with multiple models available. LoRA training tools for image generation. RAG / vector DB (choose the best integration with OpenClaw and other tools—e.g., Qdrant, Chroma, or Milvus). Multi-agent capable dev tools / environment.
Central Portal & Usability
A single web-based portal (e.g., OpenWebUI or equivalent) for central access to all tools, easy model switching, admin controls, and live demos.
Nice-to-Have / Optional Enhancements (quote separately if interested)
Full 2-node clustering with tensor parallelism (e.g., using the open vLLM-DGX-Spark repo or Ray/NCCL). Docker Compose / lightweight Kubernetes orchestration for easy updates and portability. Monitoring dashboard (Prometheus + Grafana). NVIDIA NIM microservices for optimized inference. Any other best-practice tools you recommend for integration, speed, or flexibility.
Your Profile You're deeply familiar with these tools (or eager to dive in as an enthusiast), NVIDIA DGX systems (especially Spark/Grace Blackwell), multi-node inference (vLLM, tensor/pipeline parallelism), Docker/containerization, and API integrations. You understand VRAM/unified-memory optimization and can make everything work together smoothly. Bonus if you have experience with OpenClaw, OpenWebUI, RAG pipelines, or agent frameworks. Proper English communication (written and spoken) is a must for smooth collaboration. Timeline & Expectations We have a tight deadline — I need a stable, running environment live as soon as possible. You're completely free to experiment, test, and play around with different configurations during setup, but the priority is delivering a functional, demo-ready system quickly. Speed matters, while still maintaining quality and stability. Compensation Competitive hourly rate (fully flexible and based on your region, experience, and the exact scope) or a fixed-price project bid if preferred. I'm completely open to discussion—propose whatever rate works best for you and your location. This project serves as a test case for potential further collaboration and ongoing work if it goes well. If you're the right fit, there will be plenty of exciting follow-up opportunities. Work Style & Availability This is remote work. I am completely flexible on working hours and not EU-bound. As long as you're excellent at what you do, I'm happy to work with talent from anywhere in the world (including low-income countries—great people deliver great results everywhere). If this sounds like a good fit, reply with:
Your relevant experience (especially with DGX Spark, vLLM multi-node, or similar stacks — enthusiasts with strong practical knowledge are very welcome). Rough timeline and cost estimate (with your proposed rate). Any questions or suggested improvements.
Looking forward to building something powerful together!
MMORPG Farming Bot Development Category: Bot Development, C#, Programming, C++, Game Art, Game Design, Game Development, Game Testing, Game UI, MMORPG Budget: $30 - $250 USD
19-Apr-2026 22:01 GMT
Full-service Digital Marketing & SEO Overhaul Category: Content Strategy, Digital Marketing, Growth Hacking, Internet Marketing, Link Building, Marketing, SEO, Video Editing Budget: ₹37500 - ₹75000 INR
Dark Fantasy Concept Art Category: Character Illustration, Concept Art, Game Art, Illustration Budget: $30 - $250 USD
19-Apr-2026 21:54 GMT
Wedding Highlight Reel Editing Needed Category: After Effects, Audio Services, Sound Design, Video Editing, Video Post Editing, Video Services, Video Streaming Budget: $30 - $250 USD
