Hermes Wiki
Projects/RBC_Work

ruff ideas on projects;

  1. network validation
  • allows pre-check, post-check on cat9k, arista, sd-wan devices, and expanding more device support. We allow pre-check, then implementation (ansible automation), then post-check and then AI Analysis
  • allows software change, software upgrade, hardware changes, hardware upgrade, and also ping features
  • integrated with servicenow, so all reports and change are audit ready
  • we also support power meintenance
  1. network certification
  • complete automation of certifiying image by performing tests in lab
  • basically we take details from users (network team) in ruff format, we understand requirements, develop tests, perform tests on lab, generate report, provide to network team to validate the report manually (like human approval to mark the image as golden image or not)
  1. Aegis
  • we have sflow/netflow/ipfix data in flowiq (internal name) that collects data from switches/routers through elastiflow collector. TapAgg Data gets enrich through Extrahop, Corvil. All data made available in Elasticsearch and our solution is UI presentation shows live traffic data and also shows which app talks to which app, how much traffic is there, what impacts would be there if anything goes down or allows us to see possible blast radius.
  1. network chatops
  • a chatbot that allows to query anything related to network (whatever we have present at the moment and we are expanding the scope)
  • queries related to device search - gets all details from Datawarehouse, Extrahop, Infoblox, Netbrain, Cisco, etc. (using toolbox mcp as well)
  • queries related to collect packets - from Extrahop and Corvil
  • queries related to mcps - grafana mcp, elasticsearch mcp (for Aegis query), network-cli mcp (for network ssh commands on multiple device (no pre-check, post-check))
  1. other projects there are other solutions, they are less frequently used nowadays,
  • BGP Routes (simulated), Trace Route (between nodes), TCP Metrics (metrics and health), route table, BGP RIB (at glance), Layer 3 (topology viewer)

MORE AI INITIATIVES:

OSFI ready compliance, AI generated DR assessment, 9 ADT programs, We certify the infrastructure for the 9 ADT program behind RBC's Enterprise AI agenda, cleaning each to run safely across the bank.

3100+ in one change:

  • production, SSL certification, automation for 3100+ silver peak devices built in hours and executed in a single change.

following 29 AI initiatives.

Architecture

  1. AI enabled pattern, creation: architect, agent, plus agent skill for AI, assisted patent, creation, and digitization.

2 pattern digitization: to machine, Readable JSON 125 of 175 converted, 46 in progress enables Athena 3.0 automation.

  1. AI enabled blueprint reviews: IA architecture agent, expediting Athena certification reviews for new certs and audit uplifts.

  2. AI architecture as a code: 72% complete.

  3. AI enabled disaster recovery assessment: automated app code inventory lookup, plus AI generated disaster recovery assessments, supports OSFI.

  4. AI SDLC, end to end, agent automation: requirements -> architecture -> intra code -> compliance evidence.

  5. spec driven development: the enabler underpinning the AI SDLC.

network engineering

Network Validation, Aegis, Network certification, Network Chatops are falling under this initiative - this is where I am working/contributing

  1. net certifier agent, automated testing and certification for 20 plus platforms.

  2. PIV, AI agent: validates pre and post change state.

  3. network change plan, execution agent: summarize, summarizes, impacted apps, or infra, and auto executed, frequent backend changes.(Example, branch router code upgrade.)

  4. network assist: AI, pre-post, validate, validation to catch human change errors.

  5. network, self healing: capabilities using AI agent foundation being laid

  6. AI agents at scale: across the network operational efficiency program.

  7. agentic AI framework, or harness: fed by telemetry, net flow, fix, extrahop, Corvil, for AI ops, forward

  8. forward, networks, AI prediction, currently paused resource capacity.

Hosting/ Compute

  1. AI farm design, sizing, certification, and token costing models across on-prem, public cloud, and neo clouds.

  2. AI farm buildout in the Canadian Data Centre to support RBC's AI strategy.

  3. VMware, Private AI Foundation, which NVIDA, POC, to run the AI farm on VMs.

  4. GPU RAM, caching, GPU RAM, caching platforms for AI on the roadmap.

Database

  1. Devin, or AI coding agents, accelerate, app code, refactoring during DB migration. Auto translate queries to post-grace sequel, compatible syntax.

  2. Clickhouse, enabling lumina's LLM observability workloads.

Storage, Middleware, End User

  1. AI driven storage operations and predictive analysis, governance and cost optimization.

  2. AI enablement for web sphere, liberty, migration, POC using IBM, Bob, Claude Devin, seeking LLM approval

  3. collaboration platform, AI enablement, and ADT program support.

  4. AidenNoteTaker: collaboration, integration, Aden note taker integration with WebEx, Zoom, delivered to production. RBC assist, notepaker, integration, pilot commencing with WebEx.

Compliance

  1. Patch discovery agent, AI research agent using LLM function calling, and web search to autonomously identify the latest patch versions and associated CVEs for a given platform.

  2. Document and standard compliance. Analyzer. AI powered gap analysis for policy documents against regulatory standards, build entirely by Factory AI's Droid from a single Jira ticket

  3. Config drift, authorization agent, AI agent, pipeline using autogen, multi-agent orchestration, and service now. Integration to autonomously detect, analyze, and correlate configuration, drift, events against approved CR.

  4. process analyzer agent, translate, process maps into non-software signatures to identify what's running on servers.

Industry Reference: Igor Manassypov (Cisco Systems) — GitHub Profile Review

Reviewed: https://github.com/imanassypov?tab=repositories (Systems Engineer, Cisco Systems, Toronto). 49 public repos, heavily focused on Cisco campus network automation, assurance, and AI-adjacent tooling for Catalyst Center (formerly DNAC), Meraki, ACI/Nexus Dashboard. Strong overlap (~80-90%) with the network engineering initiatives above (Network Validation, Network Certification, Aegis, Network ChatOps).

Repos somewhat aligned with our work

  • CatalystCenter-BGP-EVPN-VXLAN — Controller-managed BGP EVPN VXLAN campus fabric on Catalyst 9000 via Catalyst Center CLI templates. Full lifecycle: declarative intent, GitOps template sync, role-aware provisioning, streaming telemetry assurance via Splunk. Direct analog to our Network Validation (pre-check → automation → post-check) pipeline, just built on Splunk assurance instead of our Elastic/Extrahop/Corvil stack (Aegis).
  • campus-bgp-evpn-splunk-assurance — Defines a telemetry subscription baseline (model-driven telemetry / YANG gRPC) for per-site Splunk dashboards on EVPN VXLAN fabrics, sourced from "Model Maps" docs (NVE, EVPN Manager, EVPN Route Statistics). Same problem space as Aegis's flow-based blast-radius/traffic visibility, but via streaming telemetry (MDT) rather than sflow/netflow/ipfix.
  • ansible-dnac-swim — Opinionated, phase-by-phase SWIM (Software Image Management) pipeline on Catalyst Center using the cisco.catalystcenter Ansible Workflow Manager collection: import → golden tagging → distribution → activation → post-validation → rollback, driven from one declarative YAML data file. Near-identical goal to our Network Certification project (golden image certification), but scoped to the upgrade/activation workflow rather than lab-based test certification.
  • network_platforms_mcp_servers — Open-source MCP servers for Catalyst Center and Meraki, with a deliberate "learning progression" (basic → foundation → multi-org) plus a core/stdio/remote(HTTP) transport split. Directly comparable to our Network ChatOps MCP toolbox approach (device search, packet queries, network-cli MCP) — worth comparing his transport/wrapper architecture against ours.
  • catalyst-center-netconf-checker — Finds devices onboarded to Catalyst Center without NETCONF (so they can't receive telemetry subscriptions), with both Python (ad-hoc/CSV) and Ansible (pipeline/auditable) implementations. A small but useful pre-check pattern — same category as our pre-check/PIV agent validating state before change.
  • ansible-meraki-firmware-upgrade and Ansible-ThousandEyesCat9K — Firmware/agent upgrade automation for Meraki and ThousandEyes-on-Cat9K, tag-based targeting, idempotent, dual Python-SDK/Ansible-collection approach. Same shape as our "software change/upgrade" scope in Network Validation, applied to different platforms.
  • CatalystCenter-Templates-Github-integration — Syncs DNAC/Catalyst Center template projects with Git-hosted content (GitOps for network config templates). Relevant to our config drift/authorization agent (#28) and pattern digitization (#2) initiatives — template-as-code is a shared building block.
  • ansible-compliance, cisco_ios_selfsigned_cert_role — Compliance and cert lifecycle automation at device scale; echoes our "3100+ silver peak devices in one change" SSL cert automation and the compliance/config-drift agents (#26-28).

Take

He's independently building smaller, narrower open-source versions of the same problem categories RBC's network engineering AI program is solving at enterprise scale (pre/post-check validation, golden-image certification, telemetry-driven assurance, MCP-based chatops, template-as-code/GitOps). His repos are useful as an external reference for tool architecture (esp. MCP transport patterns, Ansible Workflow Manager data-model design) and as a signal of where the wider Cisco/network-automation industry is heading — worth periodically re-checking for new repos.

Hermes Wiki