Data Scientist (Foundational Models) – Intelligence Engine (Agentic AI)

About SingleInterfaceAt SingleInterface, we're building an AI Retail Tech Platform for multi-location brands, helping them winlocal discovery, engagement, conversion, and measurable business outcomes.Our VisionMaking AI-driven solutions simple and accessible for hyperlocal businesses to manage complex digitalmarketing needs.Our GoalFuel growth for 10 million business locations by 2030 through advanced AI-powered frictionlessexperiences.Core ValuesCustomer First • Getting Things Done • Being Authentic • Being Finicky • Being Techurious

Role SummaryYou'll help build the Intelligence Engine: the learning layer of our platform that improves itself over timeusing data, experiments, and models. This role is for people who don't just run notebooks — they shipoutcomes.Key Expectations

• Build and improve foundational ML models powering discovery, ranking, relevance, and conversion signals.
• Curate datasets (structured and unstructured), define labeling strategies, and own feature/model iterations end-to-date.
• Run model experiments: offline evaluation, online experimentation (A/B), and rapid iteration loops.
• Design evaluation frameworks for LLMs, embeddings, retrieval, and agent decisioning, including human-in-the-loop checks where needed.
• Partner tightly with Product and Engineering to translate ambiguous problems into measurable model wins.

• Build and improve foundational ML models powering discovery, ranking, relevance, and conversion signals.
• Curate datasets (structured and unstructured), define labeling strategies, and own feature/model iterations end-to-end.
• Run model experiments: offline evaluation, online experimentation (A/B), and rapid iteration loops.
• Design evaluation frameworks for LLMs, embeddings, retrieval, and agent decisioning, including human-in-the-loop checks where needed.
• Partner tightly with Product and Engineering to translate ambiguous problems into measurable model wins.
• Model training: classical ML and deep learning (PyTorch/TensorFlow), loss functions, regularization, calibration, bias/variance trade-offs.
• Representation learning: embeddings, metric learning, retrieval, similarity search, vector databases (or equivalent).
• LLM-related workflows: fine-tuning (as applicable), prompt and retrieval strategies, evaluations, hallucination checks, guardrails.
• Ranking and personalization: learning-to-rank, recommender patterns, propensity models (bonus).
• Experimentation: strong statistical thinking, causal intuition, offline-to-online translation, metric design.
• Data fluency: SQL and Python, feature engineering, data quality checks, pipeline sanity.
• Bonus: RL or bandits (explore-exploit), multi-agent evaluation or orchestration metrics.