Dave Mille

Focusing on local LLM runtimes (Ollama, vLLM, llama.cpp), GPU inference optimization, and cost-efficient developer setups.

Austin, TX Joined on Oct 17, 2026

Work

AI Infrastructure & Local Runtime Engineer

Dave Mille

Oct 23

Benchmarking Hindsight Agent Memory: Throughput, Latency, and Semantic Recall Under Load

#ollama #opensource #ai #machinelearning

3 min read

Dave Mille

Oct 21

Notes from a 2 AM On-Call: Why Our vLLM Semantic Router Dropped 40% of Requests Under Load

#ollama #opensource #ai #machinelearning

3 min read

Dave Mille

Oct 20

vLLM PagedAttention Memory Layout: Why Our 70B Deployment Stopped OOMing

#ollama #opensource #ai #machinelearning

4 min read

Dave Mille

Oct 19

Benchmarking FunASR Streaming ASR: Throughput, Memory Overhead, and p99 Latency Under Load

#ollama #opensource #ai #machinelearning

4 min read

Dave Mille

Oct 19

Benchmarking Hugging Face Transformers Inference: Memory Overhead and Throughput Under Load

#ollama #opensource #ai #machinelearning

3 min read

Dave Mille

Oct 18

GPUStack With vLLM in Production: Pinning KV Cache Before the OOM Finds You

#ollama #opensource #ai #machinelearning

5 min read

Dave Mille

Oct 18

Notes from a 2 AM On-Call: When llama-swap's Model Reload Hit a 30s Timeout Wall

#ollama #opensource #ai #machinelearning

3 min read

Dave Mille

Oct 17

Magnitude vs Ollama: Hardware-Tuned Kernels Cut Inference Latency in Half

#ollama #opensource #ai #machinelearning

3 min read

Dave Mille

Oct 17

Running Unsloth in Docker with OpenAI-Compatible API Relay and Unbuffered SSE

#ollama #opensource #ai #machinelearning

3 min read

DEV Community

Dave Mille

Badges

Writing Debut

Skills/Languages

Benchmarking Hindsight Agent Memory: Throughput, Latency, and Semantic Recall Under Load

Notes from a 2 AM On-Call: Why Our vLLM Semantic Router Dropped 40% of Requests Under Load

vLLM PagedAttention Memory Layout: Why Our 70B Deployment Stopped OOMing

Benchmarking FunASR Streaming ASR: Throughput, Memory Overhead, and p99 Latency Under Load

Benchmarking Hugging Face Transformers Inference: Memory Overhead and Throughput Under Load

GPUStack With vLLM in Production: Pinning KV Cache Before the OOM Finds You

Notes from a 2 AM On-Call: When llama-swap's Model Reload Hit a 30s Timeout Wall

Magnitude vs Ollama: Hardware-Tuned Kernels Cut Inference Latency in Half

Running Unsloth in Docker with OpenAI-Compatible API Relay and Unbuffered SSE