Career has come full circle to Academia. Talk to me about my research on Democratizing AI through a focus on memory related AI infrastructure topics.

Very focused on solving the real problem with data center cost of working with rich data at scale using containerized workloads.

1-3 year contracts, retained 1-6 week engagements, and scheduled 1 hour calls in the following areas of expertise:A. Technology trending and forecastingB. Product planning and portfolio strategyC. Data Center and Edge workloads, especially Data InfrastructureD. Memory and Storage technologies, especially for Deep Learning and Analytics use casesE. Broader semiconductor, SoC, HI trends related to the aboveF. Scalable, portable, and high-performance architectures for Near-Data Processing, Memory-semantic Networking, Memory Virtualization, and PersistenceG. Design of High performance micro-services for real-time SLAs in Cloud IaaS and SaaS, Financial Services, Supply Chain and Logistics, Health and Life ScienceH. Architecting semantically-rich services such as online ontology building, enterprise knowledge graphs, deep entity resolution and co-resolution, core entity class data quality and data decoration at scale.Past customers include Convex (now HPE), QLogic (now Marvell), CATC (now Lecroy), Ektimisi Semiotics, Fusion-io (now Western Digital), Samsung Electronics, and a major Software company (current). We are also active advisers to investment firms and strategy consulting firms where we provide due diligence and market forecasts on technology companies and technology sectors. Past customers include Albany Capital, and many others who shall remain nameless.We do not do patent or expert witness work for others. Pankaj's over 100 patents and publications are online for all to see. Pankaj has contributed to open source since his work at NASA in the 1990s and has continued to advise engineers on their open source activities.

Leading a team that plans and takes to market multiple workload-optimized NAND flash SSD technologies for Data Center and Edge from Samsung Electronics / Device Solutions / Memory Business.

Member of Executive Product Planning Team.Focus is on planning and proving new technologies that will extend our product leadership.Led SmartSSD product planning.

An unusual startup re-imagining analytics, AI, and Machine Learning around a foundation that exploits recent developments in mathematics, physics, and device fabrication toward delivering powerful online services for future cars and things, at unprecedented capability and cost.

I liaise between industry, faculty, and NSF which provides seed funding for the I/UCRC, and have specific objective to help align industry-academic agendas and to get more of the major cloud companies to participate in this effort.

Began as Consulting CTO of Fusion-io (reporting to David Flynn, CEO) Oct 2012.- Consolidated diverse advanced development effort around NVMFS; presented to board Jan 2013.Appointed SVP & CTO (NYSE:FIO) April 30, 2013.- Streamlined key OEM relationships; trained in-house and OEM SEs worldwide; launched new geos; drove key product certifications collaboratively with CIO; started strategic initiatives with the largest customer.Following acquisition by SanDisk, appointed VP & Senior Fellow July, 2014.- Founded and chaired SanDisk Technology Council, the fellowship of Fellows. Instituted Assumptions and Projections, and other processes to create a technology tracking and trending radar for product management.- Established SanDisk CTO Office with SVP Conley; led work on the company's first (2020) 5-year technology outlook.- Began iMemory (Intelligent Memory project) with key partner.Following acquisition by Western Digital, continued as VP & Senior Fellow.- Delivered 2021 5-year technology outlook (Data Center and Enterprise) for the broader company- Reoriented iMemory project with new staff and startup partners in database and semiconductor industries- Filed key IPs on new techniques for memory management and application-aware error correcting codes- Presented a keynote at MPSoC 2017 conference on Software Defined Hardware in France; upcoming presentation at EDPS 2017.- Presented persistent memory and data-centric computer architecture courses as part of SanDisk University, at industry events (HiPC Bangalore and Huawei Storage Summit) and in academia (speeches at Stanford, Berkeley, CMU, UW Seattle, UC Santa Cruz CE Department, UCSD)- Presented five year technology outlook learning at internal meetings (C-Staff, Core Strategy, Strategy Network, Product Technology Alignment) and externally at key customers (Apple, Google, FaceBook).- Supported Sales in US, EMEA and China with technology roadmap and trends presentations to key customers.

Lectures1. Beyond Search: 5 steps to insight, Kno.e.sis Center, Oct 21, 2009.http://knoesis.org/aboutus/visitorsPh.D. committee1. Christopher Thomas (graduated 2012) thesis titled "Knowledge Acquisition in a System" (solves the problem of relationship extraction)knoesis.wright.edu/students/.../Growing_fields_of_interest.pdf

Guest ed., Mar/Apr 2012 special issue on Beyond Search: Context-Aware Computing topped the list of free and paid article downloads in 2012Guest ed. (with George Pallis, et al), Sep/Oct 2009 special issue on Cloud ComputingGuest ed., August 2010 IEEE Computing Now Context Aware Computing themeGuest ed., Sep 2002, special issue on Global Deployment of Data Centers

I co-founded Whodini in August 2010. I was responsible for Whodini's technology stack and funnel.

My role was to mentor young researchers, to lead in writing proposals and getting demonstrators engineered, and to shape Labs research in close collaboration with HP's CTOs and customer-facing account teams.

I led the Live Information Management project at HP Labs, where Topher, Roger and I created key elements of an Information Routing service along with RSS syndication startup KnowNow. The Taxonom.com incubation came out of the work on automated ontology building with Topher Thomas. Taxonom2 entity search was developed as a way to automatically construct the A-Box portion of ontologies.

I helped create the Information Services and Process Innovation Lab at HP in 2005 based on an earlier collaboration with Susie Wee and Russ Daniels. A number of information management research themes were launched during this period under my technology leadership. My own research focus remained on large-scale information management. In 2006, I worked closely with the HP Software business in crafting their Information management strategy while also starting the Live Information Management project which spawned a successful incubation in 2007 to create ontology modeling services. I led this incubation (globally across US, India and Russia) and worked directly with its customers in Financial Media Industry.

I taught ISM 158 (Business Strategy and Information Systems) at UC Santa Cruz in Spring 2010 quarter.

I was the lead architect of our Information Lifecycle Management product. I led the work on re-architecting the product to include a sub-document de-duplication engine through the design of compression-enhancing routing algorithms in collaboration with members of my team and HP Labs researchers. During this period my responsibilities included technical due diligence work for HP Software and StorageWorks.

I continued to work at Compaq's NonStop Labs, moving more squarely into industry standard servers, and contributing to the InfiniBand standard. When the prolonged acquisition of Compaq by HP was consummated in 2002, I was invited to join as a key employee. During August 2001-September 2002, I was part-time at Compaq, while I created IntelliFabric. I invented multi-fabric topologies using rotated meshes and other combinatorial constructs and went on to create RDMA-accessible persistent memory.

I co-founded IntelliFabric. Designed InvisiBand, a switch-based integrated virtualization algorithm for simultaneous virtualization of server, network, and storage resources.

I was at the small and highly productive Tandem Labs; Compaq acquired Tandem in 1997. We built scalable I/O systems -- I still remember when we booted 6 Windows NT 3.51 servers with 356 disk drives shared across 96 SAN-attached SCSI controllers in 16 remote PCI chasses shared by the cluster during early 1998, using that system to establish the best TPC-C (transaction processing performance) number ever achieved by a Windows cluster.

I contributed to the Dynamic Resource Management System effort.

I taught courses in Supercomputing, Finite Elements Method, and the introductory course on Numerical Methods. I worked with Intel to establish the Intel Technology Lab at IIT Delhi and contributed cache-friendly parallel algorithms for Fourier and Legendre transforms to our project in porting weather modeling codes to India's own Param supercomputer. I also worked with computational science folks in academia (HiPC paper) and industry to parallelize key industry kernels for BEM, simulation of water, and 3d object intersection. Much of this was proprietary and unpublished work. I led a team of 5 engineers that developed two generations of Convex's CxTrace product for their SPP line under a contract from Convex/HP to IIT's Foundation for Innovation and Technology Transfer.

This was a NASA contractor position working for Recom Technologies. I worked with the Parallel Tools Group, where I worked on the Modeling Kernel for AIMS, and owned its Fortran and C parsers. Two important journal papers (one in Software - Practice & Experience about AIMS, and one in JPDC about the quadratic (!) structure of pipelined parallel programs) came out of this work, along with conference papers at SIGMETRICS, Supercomputing, and others.

I researched real and artificial neural networks, then wrote a book titled "Artificial Neural Networks: Concepts & Theory" which was published by IEEE Computer Society Press in 1992.