описание
LERANTS Technologies is building AYGIGEN, a decision intelligence system for commercial fruit and nut orchards, grounded in a professional orchard-management model. It helps orchard owners and agronomists organize orchard information, develop management programmes, and make better-founded decisions throughout the season. The MVP is under active development. Pilot preparation and onboarding begin in early 2027, followed by full-season field validation in selected commercial orchards from spring.
задачи
- Research models and methods for each feature, and define its task, expected outputs and scope.
- Prepare representative datasets, test fixtures and expert-reviewed golden sets with product and domain experts.
- Develop agents, skills and tools using the existing AI harness.
- Test feature logic and outputs, and run offline evaluations with code graders and LLM judges to measure quality and diagnose failures.
- Optimize accuracy, reliability, latency and cost through controlled experiments.
- Work with backend engineers on application integration, end-to-end testing and regression gates.
- Monitor features in production with online evaluations, turning production failures into new evaluation cases.
- Drive technical execution independently, document key decisions and incorporate feedback from domain experts and a technical advisor.
требования
- 3+ years of professional experience in ML engineering, including hands-on delivery of LLM-based applications.
- Strong Python engineering skills and experience building maintainable, tested software.
- Solid understanding of machine learning fundamentals, model evaluation and statistical reasoning.
- Ability to turn ambiguous product and domain problems into working features and make evidence-based technical decisions.
- A bachelor's degree or higher in data science, mathematics, computer science or a related field.
- Working proficiency in Armenian.
- English proficiency at B2 level or higher.
- Будет плюсом: Hands-on experience building, evaluating and monitoring LLM agents, experience with AI evaluation and observability platforms (LangSmith, Langfuse, etc)
условия
- part-time, approximately 16–24 hours per week, depending on MVP workload, with higher availability during critical delivery periods.
- Yerevan-based, working remotely from home; no office attendance required.
- December 2026, as soon as possible.