Über diese Staff AI Engineer Stelle bei Spara
About Spara
Spara is building AI agents that sell. Our agents engage buyers across chat, voice, and email, handling everything from qualifying leads to booking meetings to answering deep product questions. We work with companies ranging from fast-growing startups to large enterprises, and our product is working: customers like Rho are seeing step-change improvements in their sales pipeline within weeks of going live.
Sales technology is a massive market, and the shift to AI-native GTM motions is still early. We're building the company that defines how it plays out. We've raised $15M from Radical Ventures and Inspired Capital, backed by the founders of PyTorch and Google Cloud TPU, and Heads of Sales at Anthropic and OpenAI.
Our founders have done this before: David was co-founder and CEO of Triplemint, which he led to acquisition, and Zander was technical co-founder of Nomad Health (now post-Series D) and most recently led Platform at Hyperscience AI.
We're a 20-person team in downtown Manhattan building something that will be much bigger. Our engineering team is small and senior, and everyone has real influence on what we build and how we build it.
This Role
We're looking for someone who's been in the NLP and ML space long enough to have perspective. You've seen the field evolve, you transitioned into the LLM era with a clear understanding of what changed and what didn't, and you have strong instincts for when to prompt-engineer vs. fine-tune, how to build eval systems that actually scale, and how to make pragmatic architecture decisions under real constraints.
As a Staff AI Engineer at Spara, you will:
Go deep on the core engine that powers all of Spara's AI agents: retrieval, ranking, response generation, and the frameworks that tie them together. You'll bring a level of depth and technical judgment here that raises the bar for the whole team.
Design and scale the evaluation approach across the platform as we add new agent types and capabilities.
Bring depth on fine-tuning and training: when it's worth it, how to do it well, and how to make pragmatic tradeoffs as the platform scales.
Make foundational technical decisions that compound. At our stage, the right abstractions and architectural choices have outsized impact.
Collaborate closely with leadership to shape the product and technical roadmap. This is a role with real influence, not just execution.
Ship to production multiple times a day using a modern stack: FastAPI/Python, React/TypeScript, Postgres + pgvector, multiple foundation models (OpenAI, Anthropic), Google Cloud, Docker, GitHub Actions.
What We're Looking For
7+ years of engineering experience, with significant depth in NLP, search, retrieval, or related ML domains
Experience that predates the current LLM wave. You understand the fundamentals underneath the abstractions.
Track record of designing and deploying sophisticated ML/AI systems in production
Strong opinions on evaluation methodology. You've built eval pipelines before and know what makes them useful vs. theater.
Experience with or strong instincts around fine-tuning: when it's worth it, how to do it well, what the tradeoffs are
Ability to thrive in an early-stage environment. We're seed stage with real enterprise customers and a senior team, moving fast toward our next phase of growth.
Excellent communication skills. You'll be shaping how the team thinks about hard problems, not just solving them alone.
Nice to have:
Advanced degree in ML, NLP, or a related field
Previous startup experience, especially at companies that scaled from early to growth stage
Location & Work Schedule
Hybrid: 3 days per week in our office in downtown Manhattan (World Trade Center), with the option to work remotely on Wednesdays and Fridays.
Compensation & Benefits
Health, dental, and vision insurance
Flexible PTO
Collaborative environment with experienced engineers solving hard problems together
Real ownership over product direction and technical architecture
In-office culture in NYC with flexibility for work-life balance
