What are the responsibilities and job description for the Member of Technical Staff, Agent Engineering position at Listen Labs?
TL;DR: Listen teaches AI what people actually think and want. We're Sequoia-backed, raised $100M, and our customers include Anthropic, Google, and Cursor. We're hiring engineers who can build a complex AI-native product on a small team of former founders and top-tier builders.
Background
As AI gets better at building things, the bottleneck shifts to knowing what to build. We're the bridge between AI systems and what humans actually want. Today our customers are companies. Soon, AIs themselves will be our customers.
Our platform runs AI-moderated video interviews at massive scale. We find the right people from a network of millions, our AI conducts open-ended conversations with thousands of them in parallel, and we surface what to build next. What used to take research teams weeks per study, we do in hours.
Where it's going: every interview feeds a human preference model. We simulate human behavior at scale: how people react to new ideas, how they make decisions, how preferences shape markets, and how change ripples through society. We expose this as the Human API. An AI agent writes code, asks Listen whether users would actually want a feature, gets a grounded answer back, and iterates. Closed-loop product development at AI speed. Every coding agent will eventually need this signal.
Company Highlights
Series B with $100M raised from Sequoia, Conviction, Ribbit, AI Grant, and Pear VC. Selective team of <20 engineers including VC-backed founders, IOI medalists, and engineers from Jane Street and Tesla Autopilot. Customers include Anthropic, Cursor, Perplexity, Google, Microsoft, Robinhood, Nestlé, P&G, and Sweetgreen. Post-PMF growth: 20x year-over-year revenue. Huge market: clear path to $1M contracts at over 50% of the Fortune 2000.
Technical Challenges
Database of Humanity. Listen maintains a database of millions of people. We match profiles based on voice, face, and device IDs. Those profiles let us see how opinions change over time, prevent fraud, and find any niche audience.
Emotional Intelligence. There's a gap between what people say and what they think. Our AI interviewer reads tone, hesitation, and facial micro-expressions to go beyond the transcript. We've shipped the first version. We're working on surpassing even the best humans.
Preference Model. Updating the preference model is a research problem: what we already know, when to refresh it, which questions give the highest signal, and how to quantify the uncertainty in our predictions.
Human API. A model of millions of humans is only useful if you can call it from where decisions happen. We want to embed this into Slack, Linear, IDEs, and coding agents themselves. Imagine an agent shipping code, asking Listen what humans actually want, taking action, and iterating.
Agent Evals. Every part of our product is built AI-first. Study Composer helps customers scope and design studies. Research Agent analyzes thousands of responses and writes the report. The ceiling is what McKinsey does for $1M per engagement. The bottleneck is evaluating those qualitative outputs. Once you have the eval, you can hill-climb.
What We Look For
Background
As AI gets better at building things, the bottleneck shifts to knowing what to build. We're the bridge between AI systems and what humans actually want. Today our customers are companies. Soon, AIs themselves will be our customers.
Our platform runs AI-moderated video interviews at massive scale. We find the right people from a network of millions, our AI conducts open-ended conversations with thousands of them in parallel, and we surface what to build next. What used to take research teams weeks per study, we do in hours.
Where it's going: every interview feeds a human preference model. We simulate human behavior at scale: how people react to new ideas, how they make decisions, how preferences shape markets, and how change ripples through society. We expose this as the Human API. An AI agent writes code, asks Listen whether users would actually want a feature, gets a grounded answer back, and iterates. Closed-loop product development at AI speed. Every coding agent will eventually need this signal.
Company Highlights
Database of Humanity. Listen maintains a database of millions of people. We match profiles based on voice, face, and device IDs. Those profiles let us see how opinions change over time, prevent fraud, and find any niche audience.
Emotional Intelligence. There's a gap between what people say and what they think. Our AI interviewer reads tone, hesitation, and facial micro-expressions to go beyond the transcript. We've shipped the first version. We're working on surpassing even the best humans.
Preference Model. Updating the preference model is a research problem: what we already know, when to refresh it, which questions give the highest signal, and how to quantify the uncertainty in our predictions.
Human API. A model of millions of humans is only useful if you can call it from where decisions happen. We want to embed this into Slack, Linear, IDEs, and coding agents themselves. Imagine an agent shipping code, asking Listen what humans actually want, taking action, and iterating.
Agent Evals. Every part of our product is built AI-first. Study Composer helps customers scope and design studies. Research Agent analyzes thousands of responses and writes the report. The ceiling is what McKinsey does for $1M per engagement. The bottleneck is evaluating those qualitative outputs. Once you have the eval, you can hill-climb.
What We Look For
- You solve problems end to end. The team is split vertically, so every engineer owns a part of the product and makes decisions across the LLM pipeline, infrastructure, backend, and UX.
- You're a future or past founder. You scope your own work, think about the customers, and own your decisions.
- You care about getting things right. Moving fast is essential, but a 100% solution is much more powerful than an 80% one. When something breaks, you go to root cause.
- You're excited about pushing LLMs to their limits. We work directly with the frontier model labs on new releases and constantly probe where they break.
- You communicate complex ideas in writing. We work independently with one meeting a week, so writing is how tradeoffs, problems, and decisions get worked through together.
- You're highly technical. Most of our team started coding as teenagers and nerd out on details from language design to compilers.
- Competitive Compensation: We're backed by world-class investors, including Sequoia Capital, Ribbit, and Conviction, alongside Evantic, AI Grant, and Pear VC, and offer competitive compensation packages with meaningful equity.
- The range for this role is $180,000 – $300,000 base. Actual compensation is influenced by a wide array of factors, including but not limited to skill set, experience, and work location. If this range doesn't match your expectations, we still encourage you to apply.
- Benefits that Support You: Comprehensive healthcare and dental coverage, flexible time off to recharge, and an environment that values balance and trust.
- Room to Grow: You’ll have the opportunity to take on new responsibilities, shape processes from scratch, and grow alongside the company.
Salary : $180,000 - $300,000