Goodlight AI is building intelligent agents that automate complex workflows across retail, data, and enterprise systems. We’re looking for a Machine Learning Intern who is excited about applied AI—someone who wants to go beyond models and help ship real, production-grade agent systems.
This role is ideal for someone who enjoys working across the stack: from data pipelines and model experimentation to integrating LLMs into live products.
What you’ll do
Build and experiment with ML models and LLM-powered systems (classification, embeddings, retrieval, agents).
Work on real-world use cases such as retail personalization, workflow automation, and customer intelligence.
Design and improve prompt pipelines, evaluation frameworks, and agent behaviors.
Collaborate on data pipelines: ingestion, cleaning, feature engineering, and analysis.
Prototype and ship features quickly in a production environment.
Contribute to internal tooling for model monitoring, evaluation, and iteration.
What we’re looking for
Strong fundamentals in machine learning and statistics.
Hands-on experience with Python and ML libraries (e.g., PyTorch, scikit-learn).
Familiarity with LLMs, embeddings, vector databases, or agent frameworks is a strong plus.
Comfort working with APIs and integrating ML into applications.
Ability to move fast, experiment, and learn independently.
Clear thinking and ability to translate messy problems into structured solutions.
Bonus points
Experience building AI agents or working with tools like LangChain, OpenAI APIs, etc.
Exposure to retail, ecommerce, or customer data problems.
Experience with data engineering (SQL, pipelines, ETL).
Side projects or demos showcasing applied ML work.
What you’ll get
Direct exposure to building cutting-edge AI agents and production systems.
Opportunity to work closely with the founding team on high-impact problems.
Ownership of real features that go live to users.
Fast learning environment with high autonomy.
Logistics
Location: Remote
Duration: 3–6 months
Potential for full-time conversion based on performance
Originally posted on Himalayas