Machine learning engineer with a strong bias for Production ML engineering and MLOps.(+ relevant Data Engineering and DevOps skills)Proficient in Deep Learning techniques for Generative Modelling, attention-based (transformer) models for vision and language modeling, Graph deep learning, Deep probabilistic models, MCMC methods, Bayesian optimization, Diffusion models including quantized approaches (+flow-based generative modeling), Large Language models (including inference optimization, fine-tuning, retrieval augmented generation), Classical ML algorithms (biased to kernel methods and optimization), Time series, and sequence modeling with Deep learning and classical methods.Deep learning frameworks: Pytorch, Jax, Tensorflow, (order of preference in production and proficiency), Sklearn,+ HuggingFace (Transformers,PEFT, TRL), Adapeter ,bitsandbytes, Fastai,Distributed Training: Deepspeed (with Pytorch distributed), NVIDIA Megatron ( utilizing SLURM,K8s for clusters and orchestration ) Graph Deep learning: DGL - Pytorch backend, Pytorch geometric.MlOps Stack: Serving/inference:Ray serve, Torchserve, Triton inference server, TFx, Tf serving,Kserve, Tf lite ,ONXX , Rust native options Orchestrators/full stack workflows:Ray, Kubeflow,Airflow,Apache Beam (python API)Feast (feature store)Data Engineering / Data processing: Ray and Ray ecosystem tools, Apache Flink (JVM -Scala), Spark + Spark Streaming (Scala), Managed services from cloud providers ( AWS , GCP )( comfortable working with the Python counterparts, but they create performance bottlenecks )DevOps and Cloud: AWS, EKS, GCP, GKE, Docker, Kubernetes, Terraform + Ansible + Jenkins, Argo CD for k8s, Prometheus + Grafana (monitoring), IstioModel optimization: Nvidia TensorRT + Apache TVM, Nvidia Triton, Torch Script, Tensorflow XLA, Model compression, and Quantization techniques, understanding of Machine Learning compilationCUDA (C++), Nsight, Visual Profiler, CUDA GraphsProfiling, monitoring, and optimizing for performance: with Nsight Systems, Nsight Compute, Nsight DL Designer, Data Center GPU Monitor (DCGM), SLURMOpenai Tritonproficient in shell scripting, Linux internals, filesystems, tracing, and debugging in Linux, Git.More about me :Strong advocate of multi cloud , hybrid cloud use cases . First principles >> sticking to a tool . Strong advocate of reducing vendor dependency.Open to : ML infra engineering, ML optimization engineer,ML Engineer , MLOPS , Senior AI - Cloud solutions architect ( as long as we can run kubernetes ; I'm good)