Syntropic (thesyntropic.com)
Founding CTO / GPU Systems Engineer | Remote (US)
hn · Remote (US)
Jose has strong full-stack and infrastructure experience with proven ability to ship complex AI products end-to-end, including real-time systems and performance optimization. However, the role demands deep GPU kernel engineering expertise (CUDA/Triton fusion kernels, quantization primitives, memory hierarchy optimization) which is not evident in his CV—his ML experience focuses on model fine-tuning and LLM orchestration rather than systems-level GPU programming. The founding engineer background is valuable for startup pace, but this is explicitly a GPU systems engineering role requiring specialist skills Jose hasn't demonstrated.
Hook: Your experience shipping real-time AI systems at scale and optimizing cost-per-inference aligns with Syntropic's focus on compression and efficiency, though you'd need to rapidly develop deep CUDA/Triton kernel expertise for the core technical work.
Missing: CUDA kernel development, Triton programming, GPU memory hierarchy optimization, Quantization kernel implementation, Low-level GPU systems programming, CUDA performance profiling and optimization
Posting
Syntropic (thesyntropic.com) | Founding CTO / GPU Systems Engineer | Remote (US) | Equity-heavy We compress LLM KV cache. Measured: 8× at 3-bit with 0.43% quality delta (GPU-validated); 22× live end-to-end on a 5-tenant workload, GPU-verified this week (quantization × prefix dedup, decomposition always stated); 16.89× tenant density on a single V100-32GB. Validated across Mistral/Qwen/Llama with zero re-tuning. 90+ patents pending; design-partner pipeline; seed round in motion. The gap: our throughput path is a reference implementation. First job is the fused Triton/CUDA dequant+attention kernel — turning a validated 22× memory result into production speed. Then you own the serving stack. Process: call → NDA → paid 1–2 week trial on the real kernel → decide together. We report negative results straight; you should want to work somewhere that does. Contact: josef@podosai.com (Yosef Elimlich, founder — PODOS AI is our affiliated hardware company; the compression IP is Syntropic)