Subhash Polisetti

Intersection of Factory/Energy/MLops infra teams (supporting 👽alien dreadnought factories 🛸build-out🤖)

tldr: Language Models inference! (leading ForbesAI 50 startup from ex-pytorch🗽 team ) mostly worked on serverless/on-demand/reserved compute capacity juggle (for AI inference & training) across multiple hw accelerators & (neo)clouds(SLAs) - unix time-sharing systems still haunts even today..everything feels like a scheduling/systems-utilization problem. Richard S Sutton was right http://www.incompleteideas.net/IncIdeas/BitterLesson.html took some time to land for me ..but it did! [FLOPs go brrr.. RIP Von neumann arch 🧠🦄🚀 set gpus & myself on 🔥 fire!]

Bit of All things Security. Automation on FreeBSD hosts across multiple Data Centers worldwide with help of libraries in Python, Golang and Rust.

Infrastructure Customer Engineering (ICE) Nokia Container Services (NCS - k8s based) and Cloudband Infrastructure Software (CBIS - openstack based) Contributed to the delivery and care of above two products for telco customer applications (VNF/CNF) Infra automation and maintenance of 10000+ telecom(5G) customer servers/nodes . Some of the tools i was involved with: * Storage: ceph csi (cephrbd,cephfs), hp primera csi, nfs csi, RAID configs * Hardware: IPMI ,HPE ilo, Dell iDRAC, Nvidia Mellanox , Spine-leaf rack architecture. Nokia AirFrame Rackmount servers & wbx switches. * Networking: CNI (calico, multus-cni), ECMP routing, SRIOV/DPDK * Logging and Monitoring: zabbix(agent scripts/snmp traps), elk, prometheus ,grafana, fluentd, istio * Security: Selinux, FIPS, falco, gatekeeper OPA * Troubleshooting and discussions of cnf/vnf application helm charts, upgrade best practices , design choices & custom internal CRDs developed by Nokia. * Setting up special purpose nodes/hostgroups like SRIOV/DPDK hosting various telco workloads. * Helping them with dimensioning resources and capacity planning for new tools deployed on the cluster. * Django framework backed UI for Life cycle management of k8s and openstack (UI stability) * Internal cluster health check and validation tools development for improved Root Cause Analysis. Cloud RnD Business Unit (~350folks) sold/merged with RedHat :)

Worked on ceph, a unified distributed software defined storage system which has block-file-object level access from single managed cluster on backend. * Contributed to multiple major & minor releases of RHCEPH 2,3 by automating critical test cases in rados, iscsi & file system components. * Automating IOPS testing by including them as part of the sanity suites with tools like dd, fio having different parameters covering stress/load/idle scenarios. * Writing test cases for all possible workloads and components of rhel/ceph & various new storage features keeping regressions in mind. * Implemented container verification pipeline by configuring RabbitMQ for new containers images built for multiple parallel releases of rhceph & rhel; Integrated layered product interop testing of early rhel7.x,8.x builds with ceph. * Actively collaborated in QE-CI initiative efforts across Red Hat’s global QE organization, building common best practice CI/CD pipelines and tools from scratch. This led to the CephQE team achieving a 95% implementation of CI best practices. * Enhanced test-metric analysis for each release via integration of reportportal.io ML dashboard into automation framework. * Core maintainer of CephCI and teuthology test framework github repositories which are tightly coupled with pytest, apache-libcloud/gevent & ssh paramiko libraries, where containerized ceph is deployed on top of auto provisioned baremetal/VMs. * Analyzed and filed critical bugs in librados and ceph-rados gateway (rgw) component which is a REST interface to Ceph cluster that provides object storage with Amazon S3 and OpenStack Swift API compatibility. Acquired by IBM in 2019 :)

Specialization: ML course work: - Enterprise distributed systems (exploiting lineartime at time crystals or planck scale) - System software (forward pass of deterministic compilers and neural nets) - Advanced computer Arch (MIPS, ARM) - Microprocesor design (played with LPC1769 LPCXpresso board, NXP's ARM Cortex-M3 microcontroller, bunch of 2D vector graphics tree simulation algorithms) - Advanced Data mining (dimensionality and latest space extraction) - Testing Intelligent systems (AI models perf benchmarks) - Business Intelligence (traceroute Information theory) - Machine Learning (unsupervised and RL) - Deep Learning (neural network archs) - Recommendation systems (more of collab matrix and linear algebra)

spent time playing video games (pcmasterrace!!) and tinkering with linux distros