# 30 Telecom Engineer Interview Questions for 2026
Interviews at major operators and vendors have shifted hard in 2024-2026. SA core, cloud-native NFs, O-RAN integration, and AI/ML in RAN are now standard topics — even for roles that aren't titled around them. The questions below are sourced from active hiring loops at Verizon, Vodafone, Reliance Jio, Ericsson, Nokia, Mavenir, and several MEA/APAC operators in late 2025 and Q1 2026.
Model answers reflect what hiring managers told me they're actually looking for. Not textbook answers — interview answers.
Junior Level (0-3 years)
1. Walk me through what happens when a UE attaches to a 5G SA network.
Good answer: PRACH preamble, RAR, RRC Setup Request, RRC Setup, RRC Setup Complete carrying NAS Registration Request. AMF receives Registration Request, authenticates via AUSF/UDM (5G-AKA), establishes security context, allocates 5G-GUTI. PDU session establishment follows separately, involving SMF and UPF selection.
Red flag: confusing 4G EPS attach with 5G registration.
2. What's the difference between SA and NSA?
Good answer: NSA uses 5G NR for user plane only, with 4G EPC core and 4G LTE as anchor for control plane (Option 3x most common). SA uses 5G NR end-to-end with a 5G core (SBA architecture). NSA was the fast-to-market path; SA unlocks slicing, edge integration, and proper URLLC.
3. What are the main 5G NFs in the SBA core?
Good answer: AMF (access and mobility), SMF (session management), UPF (user plane), AUSF (authentication), UDM (subscriber data), PCF (policy), NSSF (slice selection), NRF (NF registry), NEF (exposure), CHF (charging). Bonus points for mentioning UDR vs UDM separation, BSF for binding, and SCP for service mesh.
4. Explain the 5G QoS framework.
Good answer: QoS flows identified by QFI. Each flow has a 5QI mapping to standardized characteristics (priority, packet delay budget, error rate, resource type GBR/non-GBR/delay-critical-GBR). UE-to-UPF mapping handled by SDAP at NR PDCP. Distinguish from 4G EPS bearer model — flows are inside PDU sessions, not bearers.
5. What is RRC and what states does a UE go through?
Good answer: Radio Resource Control. States: RRC_IDLE, RRC_CONNECTED, RRC_INACTIVE. INACTIVE is new in 5G — UE keeps context but doesn't have active radio resources, faster transition back to CONNECTED than from IDLE.
6. Describe SUPI vs SUCI.
Good answer: SUPI is the permanent identifier (IMSI-format or NAI-format). SUCI is its concealed form using ECIES with the home network public key. Concealment happens in the UE before transmission. AUSF/UDM decrypts. Designed to defeat IMSI catchers.
7. What's the difference between BWP and CC?
Good answer: Component Carrier is the physical carrier (e.g., 100 MHz at 3.5 GHz). Bandwidth Part is a logical sub-band within a CC, configured per UE. BWPs allow power-saving (UE operates in narrow BWP when idle-active) and varied numerologies within a CC.
8. Why is SCS (subcarrier spacing) configurable in 5G NR but fixed in LTE?
Good answer: 5G operates from sub-6 GHz to mmWave. Different bands have different phase noise, Doppler, and latency requirements. 15/30/60/120/240 kHz SCS lets the same air interface scale across FR1 and FR2.
Mid Level (3-7 years)
9. How does network slicing actually work end-to-end?
Good answer: UE registers with Requested NSSAI. AMF queries NSSF for allowed/configured NSSAI. NSSF returns slice instances. Each slice has dedicated SMF/UPF (or shared, depending on isolation level). Transport network maps slice traffic via VLAN, VPN, or hard slicing (FlexE). RAN slicing typically logical (RB sharing with isolation policies) or hard via dedicated cells.
Red flag: treating slicing as purely a core concept — interviewers want to hear RAN and transport included.
10. Walk me through a UE handover in 5G NR.
Good answer: Source gNB triggers based on measurement report (A3/A5 events). Decides target cell. Sends Handover Request to target. Target admits, returns Handover Request Ack with config. Source sends RRC Reconfiguration to UE with target config. UE detaches from source, performs RACH on target, sends RRC Reconfiguration Complete. Path switch to UPF. Source releases. Bonus: discuss Xn-based vs N2-based handover, conditional handover, DAPS for 0ms interruption.
11. What's the role of the UPF and how do you scale it?
Good answer: UPF handles user plane forwarding, QoS enforcement, traffic detection, and lawful intercept. Scaled horizontally (multiple UPFs for different DNNs/slices/regions) and vertically (DPDK/SmartNIC offload for high throughput). Edge UPFs deployed near MEC for low latency.
12. Explain how 5G handles authentication differently from 4G.
Good answer: 5G-AKA or EAP-AKA' instead of EPS-AKA. SUCI replaces IMSI on air. Anchor key (KSEAF) separates serving network from home network. Home network controls authentication via AUSF. SUPI delivered to AMF only after successful authentication.
13. What's the difference between gNB-CU, gNB-DU, and gNB-RU?
Good answer: Functional split per 3GPP. CU (centralized unit) handles RRC and PDCP. DU (distributed unit) handles RLC, MAC, high-PHY. RU (radio unit) handles low-PHY and RF. F1 between CU-DU. CU further split into CU-CP and CU-UP via E1. Bonus: O-RAN 7-2x split for fronthaul.
14. How does 5G handle URLLC?
Good answer: Multiple mechanisms: shorter slot durations (mini-slots), higher SCS, grant-free uplink, PDCP duplication across legs, dedicated 5QIs for delay-critical GBR, preemption of eMBB. End-to-end latency budgeting requires RAN, transport, and core coordination.
15. What's CSI-RS and why does it matter for massive MIMO?
Good answer: Channel State Information Reference Signal. UE measures, reports CQI/RI/PMI/LI back to gNB. With massive MIMO and large codebooks (Type II), accurate CSI feedback is the bottleneck for spectral efficiency. Bonus: discuss CSI compression, AI/ML-based CSI feedback being studied in Rel-19.
16. How would you debug a UE that registers but can't establish a PDU session?
Good answer: Walk through registration succeeded → PDU session establishment failed. Check: AMF logs for PDU session establishment request, SMF for session creation rejection cause, UPF reachability, DNN/S-NSSAI authorization in UDM, IP pool availability, policy decisions from PCF. Actual diagnostic flow, not a textbook list.
17. What's the difference between NSSF and NRF?
Good answer: NSSF selects which slice instance(s) a UE should use based on subscription and policy. NRF is a service registry — NFs register, other NFs discover. Different functions, often confused.
18. Describe how cloud-native 5G core differs from VNF-based.
Good answer: Containers vs VMs. Microservices per NF function (SMF split into multiple services). Service mesh (Istio, Linkerd) for inter-NF communication. Helm charts for deployment. CI/CD pipelines. Stateful workloads handled via operators and StatefulSets. Observability via OpenTelemetry. Resilience via Kubernetes primitives (HPA, PDB) rather than active/standby pairs.
19. What's the role of the SCP?
Good answer: Service Communication Proxy. Acts as a service mesh proxy between NFs. Handles routing, load balancing, message transformation, and increasingly mTLS. Optional in 3GPP but practically required at scale because direct NF-to-NF mesh doesn't scale.
Senior Level (7+ years)
20. How would you architect a private 5G network for a manufacturing site?
Good answer: Spectrum decision (CBRS in US, n78 dedicated in Europe, n79 in some APAC). On-prem 5G core (cloud-native, single NF cluster). Edge UPF on-site for low latency. Local breakout for OT traffic. Slicing for OT/IT separation. Time-sensitive networking (TSN) integration. Lifecycle: pre-deployment RF planning, parallel testing with existing Wi-Fi, phased migration. Discuss IT/OT integration, security zones, and vendor selection criteria.
21. Walk me through how you'd troubleshoot poor uplink throughput in a specific cell.
Good answer: Layered. Coverage check (RSRP/SINR distribution from PM counters). Uplink interference (RSSI noise floor over time). PUSCH BLER trends. PRACH success rate. UE capability distribution. Backhaul saturation. Bonus: discuss CCO (capacity and coverage optimization), MIMO rank distribution, and whether the issue is cell-edge or cell-wide.
22. How would you design a multi-vendor RAN with O-RAN principles?
Good answer: Open fronthaul (O-RAN 7-2x). DU/CU from different vendors connected via standardized F1. SMO with rApps for non-RT RIC. Near-RT RIC with xApps for sub-second control. RAN intelligent functions decoupled from vendor-specific code. Discuss interoperability testing pain points: O-RU/O-DU integration is the hard part in practice.
23. What's your view on AI/ML in RAN?
Good answer: Practical: ML for anomaly detection (cell outage), traffic prediction, energy optimization (cell sleep), beam management. Standardized: 3GPP Rel-18/19 AI/ML for NR includes CSI compression, beam prediction, positioning. RIC apps for closed-loop control. Honest about limitations: data quality, generalization across sites, operational maturity gap.
24. How do you think about energy efficiency in 5G RAN?
Good answer: Major levers: cell shutdown during low-load periods, micro-sleep, deep sleep modes, MIMO layer adaptation, carrier shutdown for under-utilized FR2. Trade-offs: KPI degradation during sleep, wake-up latency. Operator-level: fleet-wide AI-driven scheduling. Hardware: latest BBU and RU generations cut consumption 30-50%.
25. How do you handle backward compatibility when upgrading 5G core releases?
Good answer: Service-based interfaces use OpenAPI versioning. Multiple API versions in parallel during transition. Canary deployments for new NF versions. Schema migrations for state stores. Specifically: NF versioning declared in NRF profile, consumers select compatible versions. Hard part is the long tail of NF combinations across deployment.
26. Describe an outage you owned end-to-end.
Behavioral. Hiring managers want: clear timeline, what failed, what you did, what root cause was, what changed afterward. Red flags: blame, no root cause, no follow-up changes.
27. How do you stay current technically?
Behavioral. Good answers reference specific 3GPP releases, conferences (MWC, OFC, IETF), specific blogs or newsletters, and a habit of reading specifications.
28. Tell me about a time you disagreed with a senior architect.
Behavioral. Good answers show technical reasoning, willingness to be overruled, ability to commit after the decision. Bad answers show ego or passivity.
29. How do you balance vendor recommendations against your own judgment?
Behavioral. Vendors optimize for their offer. Your judgment optimizes for your network. Good answers show specific examples where you tested vendor claims independently.
30. Where do you think 5G is going and what would you bet on for the next 5 years?
Open-ended. Hiring managers test technical opinions and forward-looking thinking. Strong answers in 2026: AI-RAN, NaaS monetization specifically via CAMARA, FR3 for 6G coverage, slice-as-a-service for verticals, JCS as long-tail capability. Weak answers regurgitate vendor marketing.
Red Flags Hiring Managers Look For
From interviews with hiring managers across major operators and vendors:
- Vague answers. "There's beamforming and stuff." Specificity matters.
- Buzzword stacking without depth. Saying "AI-driven cloud-native zero-trust slicing" without being able to define each term.
- Inability to draw a system diagram on the whiteboard. If you can't sketch SBA architecture, you don't understand it.
- Treating 4G concepts as 5G. EPS bearers, SGW/PGW vocabulary in 5G context.
- No production experience signals. Junior is fine; pretending mid-level expertise without it is not.
- No questions for the interviewer. Curiosity matters.
- Bad behavioral answers. Blame, no root cause, no learnings.
How to Prepare
- Build a study sheet of the 8 interfaces (N1, N2, N3, N4, N6, N9, N32, F1, E1, Xn) and what flows on each. Hiring managers ask interface questions constantly.
- Sketch the SBA from memory until you can do it in 90 seconds.
- Pick three real outages from your career and rehearse the timeline.
- Read 3GPP TS 23.501 sections 4 and 5. Yes, the spec. Reading it differentiates you from candidates who only read summary articles.
- For O-RAN roles, lab work matters. OAI, srsRAN, O-RAN SC. Have something to show.
What the interview will actually ask you to do
The questions that separate candidates are not definitions. They are "walk me
through a registration", "this handover is failing, where do you look", "which
5QI would you use and why". All three are practice, not recall.
Free with an account: registration failure,
RACH failure and
IMS SIP 401 loop — three real captures where
something has gone wrong and you have to say what.
For structured preparation, the certification tracks cover
the procedures these questions come from, and the
practice exams are the closest thing to being asked cold.