Making AI model deployments observable at LinkedIn.

IRP runs the infrastructure behind search and ranking. When model rollouts fail here, search quality degrades for members.

As an observability lead, I’m building the substrate that explains what happened and why: persistent deployment timelines for multi-container (K8s) stateful apps (typically inference and ingestion container clusters), a context API designed for both dashboards and AI agents, and the telemetry layer powering alert intelligence.

Humans debug it today. Agents will debug it soon.

Part of team developing and supporting features for Oracle's first homegrown Infiniband product line (EDR 100Gbps): SDN software stack, host drivers, adapters and switches.

Primary liaison to all internal teams that integrate EDR and FDR Infiniband on processors (Sonoma), systems and applications (Exadata, RAC).

'NM2 Toolbox' (Python): Architected and mentored a small team to build a distributed web framework (RESTful) to manage large Infiniband fabrics with inherent high availability and user alerts built in for data center automation.

'Infiniband product liaison': Continue to work with various product groups within Oracle to design and architect Infiniband solutions for bare metal and virtual environments (SRIOV, Oracle Linux, OVM, Xen, Solaris, LDOMs)

'CollabGrid' : Spearheading a team that designs and develops a highly interactive data analysis tool, which enables engineers and senior management to visualize, personalize and share matrices on per project basis and generate human readable reports. (Python/ Django/ MySql/ Javascript/ JQuery).

• Drive integration of Infiniband product line in enterprise systems specializing in Exa family of products. Provide technical consultation and hands on support for customer-centric solutions by acting as a liaison between core engineering teams (Dev+test) and outbound enterprise teams.

• Hands on experience in developing software features focusing on Infiniband technologies used in Oracle's flagship solutions and High Performance Computing systems

*** Lead architect and developer for 'Fabric Monitor', a live

network monitoring tool that is served by a light weight web server running on embedded CentOS - part of Infiniband SUN DCS 36P Switch and Gateway products, which are shipped with all enterprise rack solutions today - Exadata, Exalogic, Sparc Supercluster, Big Data Appliance and more.

Public links to the tool:

1) http://docs.oracle.com/cd/E19671-01/835-0793-03/z40000131973869.html#scrolltoc

2) http://docs.oracle.com/cd/E19197-01/835-0784-05/z400017c1985573.html

3) Blog ref: http://blogs.oracle.com/miker/entry/how_to_monitor_the_bandwidth

• Provide direction and guidance to Infiniband core product team to improve product quality and devise solutions to assist engineering activities such as collaboration tools, test automaton frameworks, test strategies etc.

• Provide technical training to sales, field and operation teams worldwide on Infiniband technologies.

• Skills:

. Network technologies - Infiniband stack (core expertise), TCP/IP, Linux/Solaris network and system administration - kernel and IO driver internals)

. Database technologies: MySQL, SQLlite, SQL, schema design and migration (Legacy to Django framework using South - Additionally wrote stored procedures and triggers to modify schema).

. Web - GWT, Django, JQuery, Javascript, CSS3.

. Programming - Java, Python, Perl, C, Tcl, shell scripts.

. Systems - x64 and sparc platforms (system bringup from ground up)

• Held roles as lead developer and test lead for Infiniband software & hardware product line.

• Analyze software requirements to determine feasibility of design within the time/cost constraints, reporting results to senior management.

• Perform transfer of information to other engineer groups on Infiniband product line and automation suites that fit product test group's needs.

• Prepare software engineering design, specification, coding and/or documentation for developing Oracle unique Infiniband products.

• Test lead and developer for Infiniband switch product line at Sun.

• Utilize Web 2.0 technologies, Java/JSP, Algorithms/Data Structures and Database Design and Development.