Founding AI Systems Engineer, Yask

Gitwit•Bentonville, AR

17d•$170,000 - $195,000•Onsite

About The Position

Yask is building an AI system to address inefficiencies in physical operations such as stores, warehouses, and distribution routes. The system aims to translate real-world observations and business data into actionable steps, ensuring the right tasks are performed by the right people at the right time, and verifying completion. This requires an adaptive system capable of learning from real-world signals, maintaining context, orchestrating decisions, and improving over time. The role involves building this system from the ground up, starting with a live design partner in retail and distribution, and applying cutting-edge AI to solve tangible operational problems.

Requirements

Built and shipped agentic or LLM-driven systems in production.
Understand memory, context engineering, and RAG from experience, not theory.
Taken a system from prototype to production before.
Comfortable making pragmatic trade-offs to ship and learn quickly.
Like working directly with product, researchers, and users in person.
Enjoy ambiguity when it’s paired with real ownership.
Want to be a founding engineer, not a ticket-taker.
See Bentonville as a strategic advantage, not a compromise.

Responsibilities

Audit the existing prototype and decide what to keep, what to rewrite, and how to turn it into an MVP architecture.
Design and implement agent orchestration patterns, memory/context layers, and retrieval pipelines for production use.
Build the feedback loops that let the system learn from user corrections and field signals.
Ship production-grade services: APIs, data pipelines, CI/CD, observability, and security fundamentals.
Instrument telemetry so we can see usage quality, drift, task success, and system performance.
Run fast experiments tied to real user behavior and measurable improvement.
Document architecture decisions and create onboarding artifacts for future engineers.
Work in-person with product, research, leadership, and customers to ground technical choices in field reality.
Watch how work actually happens.
See where execution breaks.
Design agentic systems, memory layers, and feedback loops that make that work easier.
Ship quickly, measure impact, and make the system sharper from real usage.