Fireworks

Fireworks lets you swap out expensive closed-model APIs and route traffic to the best open or proprietary models for each task. Cut your AI inference costs by 50 to 75% without rewriting code—they're literally a drop-in replacement. Built by the PyTorch core crew (Lin Qiao was Head of PyTorch at Meta, plus engineers who shipped PyTorch compiler, ranking, and core ML infrastructure). They handle both serverless and on-demand inference, fine-tuning, and they're pushing state-of-the-art model optimization. For AI and ML engineers in Malaysia and Singapore, this matters because it's the infrastructure behind companies like Cursor, Vercel, Cognition, and Uber—when you're building production AI apps and burn rate is a real problem, you need something that actually cuts costs without sacrificing latency.

They're actively hiring for Singapore positions (Cloud Infrastructure, Applied ML Engineers, AI Field Engineers, Enterprise Account Execs), which means they're building real presence in the region. The platform gives you control over which model handles which request, so you're not locked into one vendor anymore.