AI Engineer

StarHub; Singapore; On-site; SGD 90,000 - 140,000; Full time

Job summary

StarHub is seeking an on-prem AI Engineer to merge generative AI with mobile network operations, driving zero-touch automation and autonomous operations. The role focuses on RAG pipelines, predictive analytics, and secure AI deployment within telecom ecosystems.

The ideal candidate has 2+ years in mobile networks and 1–3 years in ML/NLP, with a strong foundation in Python, ML tools, and containerized deployment. This is a hands-on, hybrid-role environment.

Qualifications

• Bachelor’s degree in CS/DS/AI or related field.
• 2+ years hands-on with mobile networks (4G/5G) and telecom protocols.
• 1–3 years in applied ML/NLP/Data Science engineering, including production contexts.

Responsibilities

• Develop GenAI & RAG pipelines and autonomous workflows using telecom telemetry and 3GPP docs.
• Engineer time-series and DL models for traffic forecasting and anomaly detection.
• Package models as containerized services (Docker/Kubernetes) and manage high-throughput serving runtimes.

Skills

Python

SQL

pandas

numpy

scikit-learn

scipy

Education

Bachelor’s in Computer Science / Data Science / AI or related quantitative field

Tools

Docker

Kubernetes

Git

MLflow

Weights & Biases (W&B)

Job description

Job Description

This on-premises AI Engineer role merges generative AI and mobile network operations to drive zero-touch automation, predictive maintenance and Autonomous operations objectives.

Core Responsibilities

• GenAI & RAG: Deploy, fine-tune (LoRA/QLoRA), LangGraph, Agentic and optimize foundational models and Telecom Language Models (TLMs). Build robust RAG pipelines and autonomous workflows using telecom telemetry (PM, CM, FM, XDRs) and 3GPP documentation.
• Predictive Analytics: Engineer time-series and deep learning models for traffic forecasting, dynamic anomaly detection, and automated root cause analysis (RCA) for network degradations.
• MLOps & Deployment: Package models as containerized microservices (Docker/Kubernetes). Manage high-throughput serving runtimes (vLLM, Triton), GPU scheduling, and experiment tracking (MLflow, W&B).
• Platform Integration & Guardrails: Interface AI workflows with enterprise OSS/BSS platforms (IBM watsonX, ServiceNow). Implement strict hallucination mitigation, output guardrails, and model drift detection.
• Cross-Domain Collaboration: Partner with Data Engineers for telemetry ingestion and collaborate with RAN/Core/Transport specialists to validate model outputs against real-world RF behaviour.
• CyberSecurity: Applying basic cybersecurity principles, ensuring data privacy, secure API endpoints and prompt security.

Qualifications

• Preferably with Network Expertise: 2+ years of hands-on experience with mobile networks (4G/LTE, 5G Core architectures, RAN, IMS, Transport, Network Management etc) and Telecom Protocols (Diameter, GTP, SIP, HTTP2, SCTP, PFCP).
• AI/ML Experience: 1–3 years in applied ML/NLP/Data Science engineering (can include production-level internships or heavy applied research).
• Education: Bachelor’s in Computer Science, Data Science, AI, or a related quantitative field.

Languages & Data

Python, SQL, pandas, numpy, scikit-learn, scipy

AI/ML & NLP

PyTorch/TensorFlow, Hugging Face, PEFT/LoRA, RAG evaluation

Vector & Big Data

Chroma, Weaviate, pgvector, PySpark, Snowflake

DevOps & Serving

Git, Docker, Kubernetes, FastAPI, vLLM, MLflow, W&B

Telecom Knowledge

RAN/EPC/5GC architecture, critical KPIs (RSRP, SINR, CQI, latency), Telecom protocols

Preferred Assets

• Experience with open-source/proprietary Telecom Language Models and datasets (3GPP, ITU).
• Familiarity with model quantization (GPTQ, AWQ, GGUF) for edge inference.
• Exposure to real-time streaming (Kafka, Flink) and data lakehouses (Iceberg).
• Understanding of telecom AI standardization (O-RAN, NWDAF).
• Active public portfolio (GitHub, Hugging Face, technical blogs).

The ideal candidate is a hybrid specialist who can comfortably debate 5G protocol stacks in one breath and optimize a vector database retrieval pipeline in the next.