GenAI Multimodal Scientist — Prompt Engineer

LTM; San Jose (CA); On-site; USD 150,000 - 230,000; Full time

Benefits offered by this job

Medical plan

Disability coverage

401(k) plan with company match

Life insurance

Paid time off

Parental leave

Job summary

LTIMindtree is seeking Applied Scientists in the United States (San Jose) to design and optimize Gen AI driven multimodal pipelines for detecting contextual moments in live video streams. The role emphasizes prompt engineering, multimodal inference, and model optimisation using Claude Nova Bedrock to maximise detection accuracy.

You will work on building pipelines for multimodal inference, video transcripts, and audio, with a focus on structured JSON metadata and integration into downstream

Qualifications

• Experience with prompt engineering for foundation models.
• Proficiency in multimodal AI systems (vision and NLP).
• Strong Python programming and data-pipeline skills.
• Experience with LLM inference and optimization.
• Familiarity with AWS cloud services (Bedrock, Lambda, S3) and analytics pipelines.

Responsibilities

• Design and optimize prompt engineering strategies for foundation models.
• Build pipelines for multimodal inference (video, transcript, audio).
• Detect and classify contextual moments and events.
• Develop reusable frameworks and moment-detection templates.
• Generate structured metadata in JSON for downstream systems.

Skills

Prompt engineering

Multimodal AI

LLM inference

Python programming

Data pipelines

Video analytics

Tools

AWS Lambda

Amazon S3

Bedrock

Job description

LTIMindtree is seeking Applied Scientists in the United States (San Jose) to design and optimize Gen AI driven multimodal pipelines for detecting contextual moments in live video streams. The role emphasizes prompt engineering, multimodal inference, and model optimisation using Claude Nova Bedrock to maximise detection accuracy.

You will work on building pipelines for multimodal inference, video transcripts, and audio, with a focus on structured JSON metadata and integration into downstream