Senior AI/ML Engineer
Transflo
- New York, New York, United States
- New York, New York, United States
About
DESCRIPTION Transflo is seeking a Senior AI/ML Engineer to lead the design, development, and continuous advancement of our Intelligent Document Processing (IDP) platform. This is a high-impact AI-first engineering role at the intersection of large language models (LLMs), computer vision, and multimodal machine learning — applied to one of the most document-intensive industries in the world.
You will architect and operate AI systems that automatically classify, extract, and interpret millions of freight documents — bills of lading, proof of delivery, rate confirmations, inspection reports, and more — with high accuracy and at production scale. You will work with foundation models, fine-tuned LLMs, and multimodal pipelines, bringing together AWS AI/ML services, modern MLOps practices, and advanced prompt engineering to push the boundaries of what automated document intelligence can do.
This role requires someone who thinks in systems: from raw document ingestion through model inference, feedback loops, retraining pipelines, and governed model deployment. AI and ML are not tools you reach for occasionally — they are the core of everything you build.
CORE AREAS OF RESPONSIBILITY AI & LLM System Design
Design and build end-to-end AI systems for intelligent document processing, combining large language models (LLMs), vision-language models (VLMs), and classical ML techniques to solve document classification, entity extraction, and data validation challenges
Architect multimodal AI pipelines that process structured, semi-structured, and unstructured documents containing mixed text, images, tables, handwriting, and complex layouts
Evaluate, select, and deploy foundation models (FMs) via AWS Bedrock, including fine-tuning, retrieval-augmented generation (RAG), and model adaptation strategies appropriate to document intelligence use cases
Develop and continuously refine advanced prompt engineering strategies — including hierarchical prompting, context-aware prompts, visual layout-aware prompts, few-shot and zero-shot techniques, multi-turn dialogue, image-text alignment prompts, and cross-attention optimization — to maximize accuracy and robustness of FM-based extraction pipelines
Stay current on frontier AI research (multimodal transformers, document foundation models, agentic LLM patterns) and translate relevant advancements into production system improvements
Machine Learning Engineering & MLOps
Design, train, and deploy scalable ML models using Amazon SageMaker, including experiment management, hyperparameter tuning, distributed training, and endpoint deployment
Own the full ML lifecycle using MLflow on AWS: experiment tracking, model versioning, artifact management, model registry, and promotion workflows from experimentation to production
Build and maintain robust MLOps infrastructure including CI/CD pipelines for model training and deployment, automated model monitoring, drift detection, and triggered retraining workflows
Optimize model inference performance and cost-efficiency using Amazon Elastic Inference, SageMaker inference optimization features, model quantization, batching strategies, and caching patterns
Implement evaluation frameworks and benchmark suites to rigorously measure model accuracy, extraction quality, latency, and regression risk across document types and edge cases
Implement and optimize multimodal ML pipelines for document classification, field extraction, layout understanding, and semantic interpretation across diverse freight and logistics document types
Integrate AWS Textract for OCR, form extraction, and table parsing; integrate Amazon Rekognition for image classification, object detection, and visual content analysis within AI workflows
Apply textual models for image classification and leverage open-source vision-language tools (e.g., LLaVA, PaddleOCR, LayoutLM variants, Donut) to extend and complement AWS-native capabilities
Design prompting and extraction strategies that account for document layout structure: bounding boxes, reading order, multi-column formats, stamps, signatures, and handwritten annotations
Serverless AI Pipelines & Platform
Build serverless AI inference and orchestration pipelines using AWS Lambda, API Gateway, and Step Functions, enabling scalable and cost-efficient document processing workflows
Collaborate with data engineers and backend platform teams to ensure clean, reliable data flows between source document ingestion, AI processing layers, and downstream data consumers
Contribute to the design of AI-powered Data as a Service (DaaS) capabilities, enabling structured, AI-extracted document data to be consumed by internal analytics platforms and external API clients
Champion observability and reliability in all AI systems: structured logging, inference latency monitoring, confidence score tracking, human-in-the-loop escalation workflows, and alerting for model degradation
Partner with data scientists, cloud engineers, product managers, and business stakeholders to align AI model capabilities with real-world document processing requirements and accuracy targets
Translate ambiguous business requirements into well-defined ML problem formulations, evaluation criteria, and iterative improvement plans
Contribute to internal AI engineering standards, reusable pipeline components, and model governance documentation
REQUIRED EXPERIENCE
5+ years of professional ML/AI engineering experience, with at least 2 years focused on LLMs, foundation models, or multimodal AI systems in production environments
Extensive hands-on experience with AWS Bedrock for deploying, prompting, and fine-tuning foundation models across multimodal and text-based applications
Deep proficiency with Amazon SageMaker for model training, hyperparameter optimization, hosted endpoint deployment, and pipeline orchestration
Proven MLOps experience with MLflow on AWS: experiment tracking, model versioning, registry workflows, and integration with CI/CD systems
Demonstrated advanced prompt engineering expertise across multiple paradigms: hierarchical prompting, context-aware and layout-aware prompting, few-shot and zero-shot learning, multi-turn dialogue, image-text alignment, and cross-attention prompt optimization
Hands-on experience with AWS Textract and Amazon Rekognition for document extraction, OCR, table detection, and image analysis within automated ML workflows
Experience building serverless AI pipeline architectures using AWS Lambda, API Gateway, and Step Functions
Working knowledge of Amazon Elastic Inference and SageMaker optimization tools for inference cost and latency management
Proficiency with AWS Deep Learning AMIs for rapid environment provisioning and reproducible ML experimentation
Strong Python skills: PyTorch or TensorFlow, Hugging Face Transformers, LangChain or LlamaIndex, and supporting data science libraries
Solid understanding of transformer architectures, attention mechanisms, tokenization, embedding models, and retrieval-augmented generation (RAG) patterns
Experience implementing CI/CD pipelines for ML systems including automated model evaluation gates, deployment promotion workflows, and rollback strategies
SKILLS/EXPERIENCE
Industry experience in document-intensive domains such as transportation, logistics, financial services, healthcare, or legal, where document accuracy and extraction quality have direct operational impact
Familiarity with transportation document types such as bills of lading, proof of delivery, rate confirmations, carrier invoices, inspection reports, or FMCSA compliance documents
Experience with document foundation models or layout-aware vision-language models such as LayoutLM, LayoutLMv3, Donut, PaddleOCR, or LLaVA
Familiarity with human-in-the-loop (HITL) feedback systems and active learning workflows for iterative model improvement using real-world production data
Experience with vector databases (Amazon OpenSearch, Pinecone, Weaviate, or pgvector) and semantic search patterns for document retrieval and RAG pipelines
Knowledge of model governance, responsible AI practices, confidence scoring, and auditability requirements for AI systems operating in regulated or high-stakes environments
Experience working in fully remote, distributed engineering team
#J-18808-Ljbffr
Languages
- English
Notice for Users
This job comes from a TieTalent partner platform. Click "Apply Now" to submit your application directly on their site.