Bjak
Machine Learning Platform Engineer
About this role
Build and operate the ML infrastructure powering A1's AI assistant platform, designing systems for model training, deployment, inference, and continuous improvement. You'll work with AI engineers and researchers to create scalable, reliable, and cost-efficient production systems that enable rapid experimentation and deployment.
What you'll do
- Design and build ML infrastructure for training, evaluation, deployment, and inference at scale
- Optimize model serving infrastructure for low-latency, high-throughput workloads
- Develop reliable data pipelines and workflow orchestration for model training and release cycles
- Create evaluation and benchmarking frameworks to measure model quality and detect regressions
- Build production observability, monitoring, and alerting systems for ML workloads
- Collaborate with AI engineers and researchers to translate model requirements into production-ready systems
What they're looking for
- Python
- PyTorch or JAX
- LLM serving frameworks (vLLM, SGLang, TensorRT-LLM)
- Distributed systems design and implementation
- Cloud infrastructure and GPU management
- ML pipeline orchestration and data engineering
- Production system reliability and observability
- System performance optimization and bottleneck analysis
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
Bjak
Bjak is a Southeast Asian fintech super app offering insurance, payments, savings, wallets, and investment products through a unified platform. The company is hiring full stack engineers, backend engineers, iOS developers, and Android engineers to build scalable features and reliable systems across mobile and web products.
- Website
- bjak.com
Likely interview questions
- Describe your experience building or operating ML infrastructure in production—what was the biggest challenge you faced?
- How have you approached optimizing model inference latency and throughput for production workloads?