AI Networking Hiring Signals — May 2026

Volume & Trend

• Total open roles (last 90d): 30 (vs previous 90d: no prior baseline — `total_roles_prev_90d` = 0 and `delta_pct` = null; treat this as the first instrumented window for `ai-networking`, not as growth).
• Distribution by level: 30/30 roles (100%) carry `level = unspecified` — zero usable seniority signal in the feed, so level-mix analysis is impossible even though raw_text plainly contains "Principal", "Staff", "Senior" titles that the pipeline is not mapping.

Top 5 Specialties (last 90d)

1. signal integrity: 5 jobs at 4 companies
2. RoCEv2: 3 jobs at 3 companies
3. network automation: 2 jobs at 2 companies
4. co-packaged optics: 2 jobs at 2 companies
5. SerDes: 2 jobs at 2 companies

Top 5 Hiring Companies (last 90d)

| Company | New Roles | Top Specialty | Avg Level | |---------|-----------|---------------|-----------| | Arista Networks | 0 | n/a | unspecified | | Cisco | 0 | n/a | unspecified | | Juniper Networks | 0 | n/a | unspecified | | Celestica | 0 | n/a | unspecified | | Credo Technology | 0 | n/a | unspecified |

Caveat: `new_roles` registers 0 for all five even though the JD sample shows Arista (5 JDs), Cisco (5), Juniper (5) and Celestica (5) actively posting. This is a company-attribution/dedup lag in the pipeline, not a hiring freeze — do not read the zeros as a slowdown.

Business-Talent Correlation Insights (3 most striking)

`correlation_links` is empty for this window (0 links). There is no business-talent pair to rank, and asserting confidence_score / gap-day numerics here would require fabricating data the input does not contain — so none are claimed. Flag for the pipeline owner: no business signals (8-K, product GA, funding) have been linked to `ai-networking` yet. When backfilled, the highest-priority pairs to reconcile are the signal-integrity (5 roles / 4 companies) and RoCEv2 (3 roles / 3 companies) clusters, because those are the physical-layer specialties that historically precede public 1.6T / lossless-Ethernet announcements.

Hiring as Leading Indicator

Per-company counts are 0 in stats (attribution lag), so clusters are stated at the pre-computed specialty level from `top_specialties`; company names below are qualitative evidence from the JD sample only.

• Signal integrity: 5 roles across 4 companies, concentrated on 112G/224G PAM4 channel work (Celestica "Signal Integrity Engineer, High-Speed Interconnect" — 224G PAM4, HFSS/ADS, BERT lab characterization). A physical-layer hiring cluster of this size precedes 1.6T platform GA; no public 1.6T launch in the data — expect 1.6T / 224G platform disclosures H2 2026.
• Co-packaged optics: 2 roles at 2 companies (Celestica "Staff HW Engineer, Optical Networking & Co-Packaged Optics" 800G/1.6T; Cisco "Staff Hardware Engineer, High-Speed SerDes and Optics"). CPO staffing leads productization by ~2–3 quarters and no shipping CPO switch is announced — expect a CPO switch NPI/announcement Q4 2026.
• RoCEv2: 3 roles at 3 companies (Cisco Silicon One + NX-OS congestion control; Juniper "Product Manager, AI-Native Data Center Fabric"). A lossless-Ethernet-for-AI talent cluster ahead of any named back-end fabric GA — expect a lossless-Ethernet AI fabric GA Q3 2026, positioned explicitly against InfiniBand.

Reverse-Hype Warnings

JD over-marketing scan: strikingly, the 20-JD sample is almost free of inflation vocabulary — no "rockstar", "world-class", "ninja", "moonshot", or "10x" appears anywhere. The over-marketing that does exist is roadmap puffery, not candidate hype: repeated "next-generation" scoping (Cisco "drive architecture of next-generation Silicon One", Juniper "AI-native", "self-driving network workflows"). The Juniper Marvis / Mist "conversational assistant" and "self-driving network" entry is the single most marketing-loaded JD and should be discounted — it describes an already-shipping product, so it inflates the "network automation" specialty (2 roles / 2 companies) with maintenance work dressed as frontier build.

Under-reported: every JD omits compensation, and 30/30 roles lose seniority even though raw_text contains explicit bars ("12+ years", "10+ years", "8+ years", "Principal", "Staff"). The pipeline is discarding parseable level data, which is why `level_distribution` is 100% unspecified. Also missing across all 20 JDs: headcount/team size, backfill-vs-new-req status, and any ship-date anchor. The Ethernet-for-AI-vs-InfiniBand competitive thesis is asserted qualitatively in the PM JDs (Cisco, Juniper) but never converted into a hiring target — so the strongest forward indicator, the physical-layer SI/SerDes/CPO cluster, surfaces only via manual specialty rollup and is invisible in the company-level stats.