Founding AI Engineer

Founding AI Engineer

Job Overview

Company: RevReply Location: Remote Department: Engineering Role Type: Full-Time

We are looking for a rare talent: someone who is genuinely excellent at AI and a top-tier engineer. You will be the most senior technical person at RevReply, reporting directly to the founder. You will own the intelligence layer end-to-end—models, evals, the data flywheel, and autonomy policy—while setting the technical direction for the platform. You will design, build, and architect the systems that power our revenue-conversion platform.

Responsibilities

The Intelligence Layer

• Data Flywheel: Wire human corrections, draft edits, and booked-meeting outcomes back into training data to create a proprietary moat.
• Calibrated Classification: Implement per-intent confidence thresholds that decide Manual/Assisted/Autopilot, governed by a shared policy stack.
• Evals as Infrastructure: Build versioned benchmarks from real traffic, locked holdouts, and rubric-based verifiers to ensure no model or prompt ships without measurement.
• Ownership Ladder: Manage shadow comparisons, canary rollouts, drift monitoring, and cost accounting; perform model refitting and post-training (SFT, preference tuning) as justified by the data.
• Reply Intelligence: Build draft generation grounded in customer context, voice profiles, and objection playbooks, utilizing human-edit data to compound quality.

Platform Architecture

• Workflow Engine: Build triggers, nested conditions, and human-approval gates to power customer revenue motions.
• Lead Routing: Develop decision-table engines for lead assignment with full audit trails.
• CRM Sync: Implement tiered write-back systems to advance pipelines safely.
• Integration Fabric: Develop a robust connector catalogue (OAuth, webhooks, idempotency, rate limits) to handle diverse sales-engagement platforms and CRMs.
• Conversion Optimization: Instrument analytics to quantify reply-to-meeting conversion metrics.

Requirements

• Deep AI Craft: Proven experience shipping LLM systems to production with strict structured outputs, custom eval harnesses, and feedback loops. Expertise in label quality, calibration, and weakness/leakage detection.
• Lifecycle Knowledge: You understand the ownership ladder (prompting → context engineering → SFT → preference tuning → RL → distillation) and can determine the appropriate solution for specific problems.
• Elite Python: 5+ years of experience shipping production-grade Python; you write code that sets the standard for the team.
• Systems Depth: Expertise in queues, idempotency, retries, webhook delivery, and relational database tuning for high-volume pipelines.
• Architecture: Ability to draw boundaries that hold, ensuring clean ports between the model layer and the product.
• Founding Mindset: Ability to make fast, reversible decisions, document processes through decision records, and set standards for future hires.
• Revenue-Domain Fluency: Understanding of sales cycles, conversion psychology, and GTM tooling.
• Tech Stack: Proficiency in our stack (Python 3.12, FastAPI, Celery, SQLAlchemy, MySQL/PostgreSQL, Azure OpenAI, Docker, Kubernetes).

Nice to Haves:

• Hands-on experience with post-training (SFT, DPO, RLVR, distillation on open-weight models).
• Experience with lead-gen ecosystem APIs or iPaaS-class integration platforms.
• Experience as a founding engineer or in early-stage startups.
• Capability to contribute to a React/TypeScript front-end when necessary.

Benefits

• High-impact leadership role at an early-stage company.
• Opportunity to own the entire intelligence layer and technical roadmap.
• Work in a "measure-don't-vibe" environment that prioritizes rigorous evaluation and explainable autonomy.

About the Company

RevReply is an AI revenue-conversion platform. We read every inbound reply across channels—email, LinkedIn, SMS, and Slack—to classify intent, draft responses under human-in-the-loop controls, route leads, advance pipeline stages, and sync with CRMs. Our platform is governed by strict autonomy policies, guardrails, and a full audit trail, ensuring that every AI decision is explainable and traceable. We are building our own intelligence engine—a provider-neutral classification system designed for high-stakes business communication.