NodJobsNodJobs

Principal Engineer, LLM

Upstart ยท United States | Remote

๐Ÿ‡บ๐Ÿ‡ธ Remote (US)Principal$238,400 โ€“ $330,200/yr4h ago

At a glance

Where
Remote โ€” open to applicants in the United States
Pay
$238,400 โ€“ $330,200/yr โ€” as listed by Upstart
Level
Principal
Experience
8+ years
Education
Bachelor's degree, or equivalent experience
Hours
Team operates on East/West coast time zones
Travel
Digital first role; most teams meet onsite once or twice per quarter for 2-4 consecutive days
Posted
November 19, 2026
Last confirmed open
53m ago
At Upstart
26 open remote roles โ€” mostly Other (6), AI & Machine Learning (4), Finance & Legal (4). Pay is listed on 26 of them ($79,000 โ€“ $330,200/yr).
How to apply
Directly on Upstart's own site โ€” no NodJobs account needed.

About the role

This role leads technical pathfinding for Upstart's Applied AI team, building 0-to-1 LLM and agent prototypes across the agent platform, borrower assistant orchestration, and internal workflows. It suits a principal-level engineer with 8+ years of software experience and hands-on background shipping non-deterministic LLM products to real users.

What you'll do

  • Rapidly prototype ambiguous bets across the agent platform, cross-agent context orchestration, borrower-facing assistant experiences, and internal workflows
  • Partner with Product and Design to validate non-deterministic UX and establish clean architectural patterns
  • Hand off golden datasets, automated evaluation harnesses, failure-mode maps, latency/cost benchmarks, and architectural blueprints to engineering squads
  • Align early with receiving squad tech leads, support joint go/no-go decisions including killing unviable bets, and stay engaged through initial productionization
  • Review early PRs and debug edge-case model behavior during productionization
  • Treat latency spikes, context loss between agents, hallucinations, and fallback paths as systems design problems
  • Make pragmatic applied AI trade-offs on model routing, fine-tuning, or distillation when latency, unit economics, or reliability require it
  • Lead by example with agentic coding workflows and verification harnesses, act as early adopter of internal tooling, and help product squads adopt higher-leverage practices
  • Build and support a reliable platform that lets non-technical users build, launch, and operate AI agents (voice/chat/text)
  • Build the orchestration layer behind a unified personal assistant that shares context across specialized agents
  • Improve internal operational workflows that directly connect to applicant and borrower experiences
  • Act as a technical pathfinder and force multiplier for the team and partner product squads, raising the applied AI bar across squads
  • Share practical requirements with internal tooling teams and elevate engineering judgment across teams through active collaboration

What you'll need

  • 8+ years of software engineering experience
  • Hands-on experience shipping non-deterministic, user-facing LLM or AI products to real users
  • Deep practical experience designing production LLM systems with structured outputs, tool calling, context engineering, retrieval/memory patterns, and multi-step or multi-agent orchestration
  • Strong distributed-systems and backend fundamentals: async orchestration, streaming UX, caching, timeouts and fallbacks, rate-limit handling, and end-to-end observability/tracing
  • Proven experience building golden datasets, automated evaluation harnesses (deterministic checks, heuristics, LLM-as-a-judge), regression gates, and failure-mode analyses
  • Instinct to use the simplest reliable tool first (deterministic logic, prompting, context engineering, guardrails) with judgment on when to route, fine-tune, or distill
  • Comfortable moving fast across backend and frontend stacks such as Python, TypeScript/React, and REST or streaming APIs
  • Track record of aligning with partner squads, handing off ownership cleanly, and partnering with Product, Design, and business leaders on latency budgets and fallback experiences
  • Bachelor's degree in Computer Science, Engineering, Mathematics, or equivalent practical experience
  • Partner directly with operations or business leaders to turn ambiguous goals into concrete latency budgets and resilient fallback experiences
  • Design non-deterministic UX and orchestration together, navigating trade-offs across latency, unit economics, reliability, and user trust

Nice to have

  • Startup or 0-to-1 incubation experience owning ambiguous problems end to end under real constraints
  • Experience designing reusable agent primitives, SDKs, or orchestration platforms that help peer teams build, evaluate, and launch AI workflows
  • Hands-on experience fine-tuning or distilling smaller models in production to improve latency, unit economics, or reliability
  • Deep daily fluency with agentic coding workflows and automated verification harnesses, and experience helping peers adopt them
  • Experience in regulated or high-trust consumer domains such as fintech, lending, or healthcare

Benefits listed by Upstart

  • Competitive base pay, bonus opportunities, and annual equity grants vesting quarterly
  • 401(k) or Group Retirement Savings Plan with company match
  • Employee Stock Purchase Plan (US only)
  • Medical, dental, vision, and wellness coverage; supplemental health coverage for Canada
  • Health Savings Account contributions for eligible plans (US only)
  • Life insurance and disability coverage
  • Paid time off, sick leave, and company holidays
  • Paid family and parental leave
  • Family-centered benefits for fertility, parenthood, and caregiving
  • Employee Assistance Program with mental health support
  • Financial wellness resources including financial planning tools and a financial concierge (US only)
  • Annual wellness allowance

Tools & skills mentioned

PythonTypeScriptReactstreaming APIsLLMtool callingretrievalLLM-as-a-judgeobservabilitytracingfine-tuningdistillation

Summarized by NodJobs from Upstart's posting. Read the full posting on Upstart's site โ†’

NodJobs is free for job seekers โ€” we never charge to apply, and this role takes you straight to the company's own application.

Similar live jobs