Senior AI Data Engineer, VideoAmp

• 4+ years of experience in AI/ML engineering, applied ML systems, or data engineering
• Demonstrated experience shipping LLM-powered systems into production
• Strong understanding of:
• Prompt engineering and evaluation frameworks
• Embeddings and similarity search on structured and semi-structured data
• Hybrid AI systems that combine LLMs with deterministic logic
• Proficiency in Python and SQL; strong software engineering fundamentals
• Experience working with large-scale data platforms and distributed systems
• Ability to design AI systems that meet measurement-grade reliability, explainability, and auditability requirements

Desirable

• VideoAmp is seeking a Senior AI Data Engineer to join our Linear Data Processing (LDP) team. This role focuses on building production-grade AI and LLM-powered systems directly into large-scale data pipelines that support audience measurement, currency, identity, and metadata workflows across linear and digital media
• You will apply LLMs and classical ML to real-world linear TV problems: network-specific schedules, hourly viewership logs, program metadata, and commercial schedules that vary widely by partner, format, and quality
• Design and operate AI- and LLM-enabled data pipelines processing billions of linear and digital records
• Build and productionize LLM-driven systems for concrete LDP problems, including:
• Network & Program Schedules
• Normalize network-specific program schedules with inconsistent naming, time boundaries, and formatting
• Infer program identity, episode, and season from sparse or ambiguous schedule metadata
• Detect and reconcile schedule discrepancies across multiple upstream providers
• Hourly Viewership Logs
• Interpret and normalize hourly (and sub-hourly) viewership logs with varying schemas and aggregation rules
• Use LLMs to assist with schema inference, validation, and semantic consistency checks
• Generate structured explanations for anomalous spikes, drops, or gaps in viewership data
• Commercial & Ad Spot Metadata
• Enrich and classify commercial schedules and ad spot metadata, including creative names, durations, and placements
• Resolve inconsistencies between planned schedules and aired spots
• Support automated deduplication and alignment of ad spots across networks and feeds
• Combine LLMs with rules, heuristics, and ML models to ensure deterministic, auditable outcomes
• Own the end-to-end AI lifecycle, including:
• Prompt design and evaluation
• Feature engineering and training pipelines
• Deployment, monitoring, drift detection, and retraining
• Integrate LLMs responsibly into batch and event-driven LDP workflows, balancing accuracy, latency, and cost
• Build human-in-the-loop workflows for high-impact or low-confidence cases
• Collaborate closely with Measurement, Identity, Library, and Product teams to translate ambiguous data issues into scalable AI solutions
• Provide technical leadership through design reviews, mentorship, and architectural guidance
• What Success Looks Like at VideoAmp:
• LLM-powered systems materially reduce manual curation of schedules and metadata
• Data quality issues are detected earlier and explained more clearly
• Measurement pipelines become more adaptive to partner-specific formats
• AI systems are trusted, observable, and operationally stable