Mohammad Waqas

Intro

I am a software developer and AI infrastructure engineer with 15+ years of software/systems experience across Linux, Python, C++, Kubernetes, NVIDIA GPU infrastructure, local LLM inference, and observability. My current focus is building practical AI infrastructure and platform-operations labs for private, local, and edge LLM deployments. I work with k3s, Helm, Ollama, Open WebUI, FastAPI, Prometheus, Grafana, OpenTelemetry, NVIDIA GPU tooling, DCGM-style metrics, llama.cpp, GGUF models, and operational runbooks. I have built open-source projects around Kubernetes-native LLM observability, GPU monitoring, local inference debugging, CUDA-first Python SDKs, Cursor/llama.cpp integration, support-bundle automation, smoke tests, benchmark workflows, and RCA-style troubleshooting documentation. My main project, llm-observability-stack, is a Kubernetes-based AI infrastructure stack for running, exposing, monitoring, and debugging GPU-backed LLM services. It includes LLM request metrics, TTFT, token throughput, p95 latency, generated tokens, active requests, error rates, GPU temperature, GPU memory, Prometheus/Grafana dashboards, OpenTelemetry traces, and runbooks for platform troubleshooting. I am also the creator of llamatelemetry, a CUDA-first Python SDK for local GGUF/llama.cpp inference observability. It provides inference-engine abstractions, llama-server lifecycle management, OpenAI-compatible client workflows, OpenTelemetry tracing, GPU runtime metrics, model tooling, benchmark utilities, and reproducible experiments on local NVIDIA hardware and Kaggle dual Tesla T4 GPUs. Earlier in my career, I worked on embedded simulation and systems software using C++ and Java, with a focus on concurrency, I/O workflows, troubleshooting, and performance optimization. I also have 7+ years of customer-facing technical support experience, helping users reproduce issues, investigate failures, review logs/configuration, document fixes, and translate user problems into engineering feedback. My engineering style is hands-on and operations-focused: reproduce the issue, inspect logs/metrics/traces, isolate configuration problems, validate fixes, document root cause, and convert lessons into repeatable runbooks. I am interested in remote EMEA / European-shift roles in AI infrastructure, Kubernetes platform operations, GPU infrastructure, observability, DevOps, SRE, technical support engineering, platform reliability, and open-source AI tooling.

Links