Job description

Software Engineer, AI New York City. In person, five days a week. $130,000 to $170,000 base, 0.3% to 0.75% equity. What Custral is Most CRMs stop at the closed deal. Everything after it, implementation, support, renewal, product feedback, ends up spread across five tools that don't share a record. Teams bend their process to fit the software and keep the real work in spreadsheets. Custral is one workspace for the whole loop. Marketing, sales, onboarding, success and product work from the same customer records. You shape the objects, views and workflows around how you run, instead of the other way around. AI agents live in the data model rather than on top of it, with real scoping, tool gating and model selection across OpenAI and Anthropic. Stella, our assistant, works the same records the rest of the product does, and asks before she sends anything. It is early. We're letting teams in a few at a time. The role Own the part of Custral that does the work on its own. Agents here are not a chat box bolted onto a CRM. They have first-class scoping, tool gating and model selection across OpenAI and Anthropic, and they write into the same records the rest of the product reads. Stella sits on top of that, and by default she asks before she sends anything. You'd own the runtime underneath both: the tool layer, how context gets assembled out of Postgres and OpenSearch, how a run is traced when it goes wrong, and how we know a change made it better instead of just different. That last part matters most. Agent quality is not a vibe. You'll build the evals, the regression suites and the tracing that let us ship a prompt or a model change on evidence. It is still full-stack work. The agent surfaces are React, the runtime is Node and Express, and you'll build both. Work like this is on the table for your first quarter: • The agent eval harness . Today a prompt or model change ships on judgment. You'd build the regression suite and tracing that lets it ship on evidence. • Stella's capabilities . New tools, better context assembly out of Postgres and OpenSearch, and more of the surfaces where she does the work instead of describing it. What we look for You'll be a fit if: • You're deeply fluent in TypeScript and comfortable across the stack, not only in the model layer • You've shipped LLM features to real users and kept them working as models changed underneath you • You've built evals or regression suites for non-deterministic systems, and you can say what they caught • You understand retrieval and context assembly as engineering problems, not as a library call • You've worked with Postgres and something search-shaped, OpenSearch or Elasticsearch • You do your own research and move work forward without waiting for direction Useful, not required : agent frameworks and MCP; streaming UIs; Redis for queues; cost and latency tuning at volume. Don't apply if your LLM experience is prompt-only, or if you want to do research rather than ship. This role is product engineering with models in it, and the thing we judge is whether a customer's day got better. Strong candidates ship real functionality without cutting corners, explain their tradeoffs, and leave the system better than they found it. How we work Deliver Impact Through Excellence . It's the one lens we hire on, and we ask about it in every interview. Impact, how we do the work: • Customer Focus . We start from what customers need and check our work against it. • Candor and Transparency . Say what you think, early, and argue about the work in the open. • Clarity and Focus . Pick the simplest thing that solves the problem, and ship it. • Embrace Change . Drop an idea that isn't working, including your own. • Think Big, Act Bold . Take on the hard problems customers actually have. • Attention to Detail . Test the edge cases and finish the details before calling it done. Excellence, how we work with each other: • Top-Tier People . People who are very good at their work and want to do it at an early-stage company. • No Ego . The best idea wins, whoever it came from. • Proactive Mindset . If you see a problem, fix it or flag it without waiting to be asked. • Future-Mindedness . Know how your work affects the rest of the company, and plan for what comes next. • Continuous Learning . Keep learning from customers, from teammates and from what went wrong. • Diversity of Thinking . Different backgrounds catch different mistakes, so we hire for range. We give people a lot of autonomy and expect them to own the outcome. We keep process to the minimum that keeps us honest. This is a five day a week in-person role in New York. At this stage, being in the room is worth more than the flexibility it costs. It's early-stage work : the problems are ambiguous, the pace is high, and there will be long days. The tradeoff is that the decisions you make in the next year compound for the life of the company. How we hire Four steps, under seven hours in all, and no take-home. We look for our values at every one, not with resume checklists or trick questions. • 15 min, phone screen . A first call to see if there's a fit on both sides. • 30 min, experience screen . Your experience, and how you think. • 60 min, role interview . A technical interview on the work of the role. • 4 hours, onsite . In person, at our New York City office. We're looking for this person to start in January of 2027. Compensation and benefits • Base salary: $130,000 to $170,000 • Equity: 0.3% to 0.75%, four year vest with a one year cliff • Medical, dental and vision, fully covered for you • A monthly wellness stipend for gym, fitness and mental health • Lunch and dinner at the office, and a commuter stipend • Unlimited PTO. Everyone takes at least three weeks. Where you land in the range depends on experience and scope, not tenure or credentials. Disclaimers This role is based in New York, NY, and requires in-person work five days per week. Relocation assistance may be available on a case-by-case basis. Custral is an equal opportunity employer. We are committed to creating an inclusive environment for all employees and applicants. We do not discriminate on the basis of race, color, religion, sex, sexual orientation, gender identity or expression, national origin, age, disability, veteran status, genetic information, marital status, or any other characteristic protected by applicable law. In compliance with New York law, the compensation range listed for this role is a good-faith estimate of the base salary range Custral expects to pay for this position at the time of posting. Actual compensation may vary based on experience, skills, performance, and role scope. Custral is committed to providing reasonable accommodations to qualified individuals with disabilities and for sincerely held religious beliefs, in accordance with applicable law. If you require an accommodation during the hiring process, please let us know.

Requirements

Must-haves

Software Engineer, AI
New York City. In person, five days a week. $130,000 to $170,000 base, 0.3% to 0.75% equity.
What Custral is
It is early. We're letting teams in a few at a time.
Own the part of Custral that does the work on its own.
That last part matters most. Agent quality is not a vibe. You'll build the evals, the regression suites and the tracing that let us ship a prompt or a model change on evidence.
It is still full-stack work. The agent surfaces are React, the runtime is Node and Express, and you'll build both.
Work like this is on the table for your first quarter:
The agent eval harness . Today a prompt or model change ships on judgment. You'd build the regression suite and tracing that lets it ship on evidence.
Stella's capabilities . New tools, better context assembly out of Postgres and OpenSearch, and more of the surfaces where she does the work instead of describing it.
What we look for
You're deeply fluent in TypeScript and comfortable across the stack, not only in the model layer
You've shipped LLM features to real users and kept them working as models changed underneath you
You've built evals or regression suites for non-deterministic systems, and you can say what they caught
You understand retrieval and context assembly as engineering problems, not as a library call
You've worked with Postgres and something search-shaped, OpenSearch or Elasticsearch
You do your own research and move work forward without waiting for direction
Useful, not required : agent frameworks and MCP; streaming UIs; Redis for queues; cost and latency tuning at volume.
Don't apply if your LLM experience is prompt-only, or if you want to do research rather than ship. This role is product engineering with models in it, and the thing we judge is whether a customer's day got better.
Strong candidates ship real functionality without cutting corners, explain their tradeoffs, and leave the system better than they found it.
Deliver Impact Through Excellence . It's the one lens we hire on, and we ask about it in every interview.
Customer Focus . We start from what customers need and check our work against it.
Candor and Transparency . Say what you think, early, and argue about the work in the open.
Clarity and Focus . Pick the simplest thing that solves the problem, and ship it.

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

airestpostgresreacttypescriptllm

Platform: dover
Effort: medium

Benefits

Benefits detected in the listing.

Dental

Vision

Pto

Wellness