Principal GenAI & Retrieval Architect

Photon; Boston (MA); On-site; USD 170,000 - 240,000; Full time

Job summary

Photon is seeking a Principal Data Scientist specializing in Generative AI to serve as the top technical authority on model science, advanced retrieval, AI quality, and evaluation strategy. You will define the scientific vision for enterprise GenAI capabilities and pioneer novel RAG architectures, evaluating when to use RAG vs.

fine-tuning and ensuring production AI systems maintain high accuracy and safety.

Qualifications

• Extreme experience in Data Science and Applied ML with leadership in Generative AI.
• Deep mastery of LLM architectures, attention mechanisms, and alignment methods (DPO/RLHF).
• Proven track record building enterprise-grade RAG architectures and vector search stacks.
• Experience evaluating AI systems for safety, accuracy, and trust in regulated environments.
• Strong Python and ML/DL framework skills (PyTorch, Scikit-learn, Pandas, NumPy).
• Experience deploying models across cloud platforms (Azure, AWS).

Responsibilities

• Define enterprise science roadmaps for Generative AI, retrieval, and AI safety.
• Pioneer RAG paradigms and evaluation frameworks across production systems.
• Lead model selection, adaptation, and optimization between proprietary and open models.
• Mentor senior data scientists and offshore teams; translate research into production-ready features.
• Represent the company as a technical authority in AI innovation to executives.

Skills

Data Science

Generative AI

LLMs

Retrieval systems

RAG architectures

Python

PyTorch

Scikit-learn

Pandas

NumPy

PEFT/LoRA

DPO/RLHF

GraphRAG

Knowledge Graphs

Vector search

Neo4j

Azure/AWS Databricks

Tools

Pinecone

Qdrant

OpenSearch

pgvector

Neo4j

Databricks/MLflow

Job description

Photon is seeking a Principal Data Scientist specializing in Generative AI to serve as the top technical authority on model science, advanced retrieval, AI quality, and evaluation strategy. You will define the scientific vision for enterprise GenAI capabilities and pioneer novel RAG architectures, evaluating when to use RAG vs.

fine-tuning and ensuring production AI systems maintain high accuracy and safety.