Second Nature Computing
Founding AI/ML Engineer
Second Nature Computing, San Francisco, California, United States, 94199
Location
San Francisco, CA (HQ) — in-person most days with flexibility when you need it. Must be authorized to work in the U.S.
About Us We believe everyday life shouldn’t have us juggling apps. Second Nature Computing is building intelligence that senses, learns, and acts in the background so the right stuff happens without you asking. Our first product,
Poppy , is a proactive consumer companion that lives on your iPhone. Poppy plugs into your calendar, email, music, fitness level, and many more integrations to nudge you with helpful suggestions throughout your day.
What we’re looking for A builder‑researcher hybrid who’s shipped LLM products in production, obsesses over latency graphs, can whip up a jaw‑dropping demo in a weekend hack, and ship it to production the following week.
What you’ll do (first 90 days)
Orchestrate LLMs to across our in-house context engine, aiming to achieve sub-500ms end‑to‑end latency.
Build eval pipelines, orchestration graphs, workflows, and kick off on‑device inference pilots.
Decide when vector DBs beat KV caches and implement the winner.
Must‑haves
Proven prod LLM track record.
Python, Torch/Ray, LangGraph/LangChain, OpenAI fine‑tuning chops.
Deep intuition for latency/accuracy/cost trade‑offs.
#J-18808-Ljbffr
About Us We believe everyday life shouldn’t have us juggling apps. Second Nature Computing is building intelligence that senses, learns, and acts in the background so the right stuff happens without you asking. Our first product,
Poppy , is a proactive consumer companion that lives on your iPhone. Poppy plugs into your calendar, email, music, fitness level, and many more integrations to nudge you with helpful suggestions throughout your day.
What we’re looking for A builder‑researcher hybrid who’s shipped LLM products in production, obsesses over latency graphs, can whip up a jaw‑dropping demo in a weekend hack, and ship it to production the following week.
What you’ll do (first 90 days)
Orchestrate LLMs to across our in-house context engine, aiming to achieve sub-500ms end‑to‑end latency.
Build eval pipelines, orchestration graphs, workflows, and kick off on‑device inference pilots.
Decide when vector DBs beat KV caches and implement the winner.
Must‑haves
Proven prod LLM track record.
Python, Torch/Ray, LangGraph/LangChain, OpenAI fine‑tuning chops.
Deep intuition for latency/accuracy/cost trade‑offs.
#J-18808-Ljbffr