DoorDash hires for production migration from frontier APIs to open-weights models

swaroopch · x · 2026-07-22

DoorDash is hiring an engineer to move real product workloads from proprietary APIs to open-weights models in production.

• The team says it has been investing in open weights for more than a year and has already had notable success.
• The role sits between infra and product and covers the full path: evals, prompt tuning, benchmarking, and inference.
• The post explicitly looks for someone who can compare sglang vs. vLLM and understands where models like Qwen, GLM, Kimi, Inkling, and Nemotron are strong or weak under production load.
• The goal is to run the same workload on a tuned open-weights model, making it cheaper and faster than a frontier-model API.

It’s a useful signal that large product teams are treating open-weights migration as a serious production engineering problem, not a hobby project.

Original post →

More from Companies & People

• Amazon cuts jobs from its AGI team — Polymarket · 2026-07-22
• DeltaNet notes unpack Kimi Delta Attention with equations and a state-update diagram — nrehiew_ · 2026-07-22
• Open models near frontier products force labs to justify what customers pay for — PeterDiamandis · 2026-07-22
• DSA2026 Kampala summer school wraps with a focus on ethical, trustworthy AI — lawrennd · 2026-07-22
• AMD and Anthropic strike a $5 billion deal for 2 gigawatts of AI compute — The Verge AI · 2026-07-22
• A fake $4,699 AI CEO replacement ad turns corporate satire into a meme — DionysianAgent · 2026-07-22

Companies & People channel →