Senior Machine Learning Founding Engineer

About the Client & Project

We are representing a high-potential AI product venture backed by a $100M+ initial commitment from a highly profitable global tech group. The company is building next-generation, AI-native workspace and communication software designed to fundamentally transform how millions of users execute daily workflows.

Rather than building simple conversational chatbots, the team focuses on Agentic AI: multi-step reasoning, tool integration, and persistent long-term context operating under permissioned autonomy, while keeping the user in control. Their flagship product shifts users from manually processing high-volume communications and daily tasks to reviewing and approving AI-completed work.

The team combines deep AI research with rapid, hands-on production engineering.

Why Join This Team?

• Founding Member Impact: Drive technical architecture, engineering culture, and product strategy from Day 1.
• Massive Financial Runway: Backed by an initial $100M investment, combining the speed and ownership of a startup with long-term financial stability.
• True Remote Autonomy: Work in a high-trust environment with full flexibility over working hours and location.
• Top-Tier Compensation: Highly competitive cash and equity structure.
• Meaningful Engineering Challenges: Solve non-trivial production issues around model non-determinism, real-world tool execution, and low-latency agent reasoning.

Core Focus & Responsibilities

• Core ML & Agent Systems: Architect and ship production-grade ML systems powering proactive AI agents, task-triage engines, and multi-step reasoning workflows.
• End-to-End ML Ownership: Take full accountability for the ML lifecycle: data curation, model fine-tuning/training, inference optimization, evaluation, and production monitoring.
• Agentic Logic & Guardrails: Build reliable systems that interact with external tools and APIs while maintaining high reliability, safety, and persistent contextual memory.
• Research to Production: Translate state-of-the-art LLM and agentic research into high-availability software handling real-world user workflows.
• Technical Leadership & Mentorship: Lead design reviews, set high engineering standards, and mentor peers across the ML team.
• Cross-Functional Ownership: Partner closely with Product, Engineering, and Research leads to translate complex user problems into shipped features.

Tech Stack & Key Concepts

• Primary Language: Python
• ML Frameworks: PyTorch / JAX
• Core Paradigms: Agentic workflows, function calling/tool integration, long-context memory, automated workflow triage
• Infrastructure: GPU-accelerated training and inference pipelines, distributed systems, evaluation and observability frameworks

Ideal Candidate Profile

• Production ML Experience: Proven track record of architecting, deploying, and maintaining production AI/ML systems at scale.
• Systems Engineering Mindset: Strong software design principles, including clean code and scalable architecture, with a focus on robust systems over one-off scripts.
• Agentic AI Intuition: Understanding of modern foundation-model behavior, failure modes, function calling, and edge-case resolution in live environments.
• High Ownership & Velocity: Comfortable navigating ambiguity, prioritizing under production constraints such as latency, cost, and safety, and driving projects independently to completion.

Key Outcomes & Expected Impact

• Delivery of core ML subsystems that consistently meet strict targets for reliability, speed, and cost-efficiency.
• Scalable data, training, and inference infrastructure designed for long-horizon context and persistent memory.
• Measurable reduction in manual user effort, delivering a seamless AI-driven product experience.

Recruitment Process

• Initial Screening Call: Recruiter / HR alignment
• Technical Deep-Dive: Discussion with Technical Leads / System Design
• Hands-on Architectural Session: Practical ML and agentic systems engineering challenges
• Final Culture & Alignment Call: Prompt decision and offer

Total of 4 stages maximum, with fast feedback at every step.