AI / MLOps & devops Engineer — building production LLMs, scaling GPU workloads, and shipping reliable cloud-native AI platforms. I lead model engineering and platform efforts at DrishtIQ-AI, driving end-to-end LLM development (LoRA / QLoRA), GPU kernel optimization (CUDA / C++), and low-latency production serving. I design and ship MLOps pipelines and cloud infrastructure that turn research into dependable, high-performance products. Key achievements Engineered a custom finance LLM (LoRA / QLoRA on LLaMA) — +76% uplift in predictive accuracy for risk analysis. Built a high-availability RAG stack (LangChain + Pinecone) powering 10+ domain models with sub-second retrieval. Modernized infra to Amazon EKS using Terraform + ArgoCD and automated CI/CD (Jenkins, GitHub Actions), cutting deployment errors by 42%. Optimized GPU kernels and implemented NVIDIA benchmarking/profiling pipelines (Nsight, nvprof), significantly accelerating training & inference. Core tech CUDA / C++ | PyTorch | Hugging Face Transformers | LoRA / QLoRA | vLLM | LangChain | Pinecone | MLflow | Docker & Kubernetes (Helm, ArgoCD) | AWS EKS | Terraform | Jenkins | GitHub Actions | SageMaker | ngrok I solve hard problems at the intersection of ML theory and production engineering — scaling models, squeezing performance from GPUs, and building resilient MLOps for real business impact. Open to senior SRE / MLOps / AI roles and collaborations.