Data Scientist – LLM Systems & Video Ads

Location:

Team:

Role Overview

As a Data Scientist at blkbox.ai, you will architect, implement, and scale data-driven and LLM-based systems that generate high-performing mobile gaming ads at production level. This role blends data science, multimodal AI, generative models, and LLM automation—focusing on retrieval pipelines, embedding-based similarity search, and production-grade model operations.

You will design and deploy end-to-end LLM workflows—from prompt design and model orchestration to RAG pipelines and vector database integration—enabling automated ad creation.

Key Responsibilities

LLM System Design, Data Intelligence & Workflow Engineering

• Architect LLM-powered workflows for narrative generation, scripting, voiceovers, storylines, and other ad concepts
• Build data-driven logic and model selection strategies informed by creative performance and mobile gaming outcomes
• Run structured data analysis using Python and SQL to identify creative attributes correlated with winning ad performance
• Integrate multimodal data sources into creative intelligence pipelines

LLM, RAG & Vector Retrieval

• Use LLM APIs such as OpenAI GPT and Google Gemini for scripted and generative workflows
• Build retrieval-augmented generation (RAG) pipelines for:
• story/voiceover generation
• gameplay retrieval
• creative asset selection
• Work with modern embedding models (OpenAI, sentence-transformers, etc.)
• Implement vector search via Pinecone, Weaviate, Chroma or FAISS
• Experiment with prompt engineering, prompt templates, function calling, and structured prompting

Production AI, Monitoring & Optimization

• Collaborate with backend engineers to integrate LLM components into production pipelines
• Monitor and optimize LLM workflows for:
• latency
• token usage and cost
• throughput
• failure rates
• hallucinations
• Evaluate multiple LLMs and embeddings for performance, relevance, and cost efficiency
• Implement reproducible workflows, model evaluation systems, and reliability checks at scale

Cross-Functional Collaboration

• Partner with production and creative teams to translate creative requirements into systematic generative pipelines
• Present technical recommendations clearly to product, engineering, and creative stakeholders
• Own experiments end-to-end including benchmarking, metrics, prototype automation, and deployment

Required Skills & Qualifications

Technical

• Strong proficiency in Python and SQL
• Hands-on experience with LLMs, RAG concepts, embeddings, and vector search
• Experience calling model APIs (OpenAI, Gemini, etc.)
• Familiarity with LangChain, LlamaIndex or similar frameworks is a plus
• Strong understanding of:
• LLM evaluation frameworks
• prompt engineering patterns
• latency and cost optimization
• model hallucination control
• Solid foundation in statistics, experimentation, and ML deployment workflows

Soft Skills & Behaviors

• Ability to communicate complex technical concepts to non-technical partners
• Curious, inventive, and motivated by experimentation and rapid iteration
• Adaptability in fast-moving production environments
• Strong ownership mindset and comfort working cross-functionally

Why This Role Is Unique

• You’ll be building real production LLM systems—not just prototypes
• You’ll work across creative, multimodal, and generative domains
• You’ll influence automated ad production for some of the largest mobile gaming companies in the world