Principal AI/ML Architect

The Opportunity

This is a senior technical client leadership role that blends deep hands-on ML expertise with strategic advisory and consulting skills. You will be the most experienced ML voice across a diverse and expanding book of customer engagements — from early-stage companies bringing ambitious ML ideas to market, to established enterprises modernizing how they build and operate AI systems on AWS.

You will shape strategy, influence architecture, and leave every team you touch better than you found it. You bring the scientific depth to design and evaluate models rigorously, the engineering depth to architect production ML systems at scale, and the consulting instincts to translate both into business value for customers.

If you have led the hard conversations, shaped the architecture decisions that mattered, and built the things others benchmark against — and you are looking to do that across a growing portfolio of varied and interesting customers — this is the role for you.

What You'll Do

• Lead end-to-end ML assessments across infrastructure, data pipelines, model lifecycle, and organizational readiness — producing recommendations that drive executive decision-making and earn Caylent the next engagement.
• Partner with sales and solutions teams through the proposal and scoping phase, contributing the technical depth needed to shape well-grounded statements of work.
• Serve as the senior technical authority on client engagements — possibly across multiple projects simultaneously — providing architectural guidance, ensuring technical quality from your project team members, and getting hands-on when the engagement demands it, without owning day-to-day implementation responsibilities.
• Own or orchestrate high-quality POCs that give customers confidence before committing to a larger initiative.
• Advise customers on ML operations standards and architecture — covering MLOps pipeline design, model lifecycle management, LLMOps patterns, and production monitoring frameworks — translating operational complexity into decisions and guardrails their teams can own and sustain.
• Shape how Caylent wins its most technically complex opportunities — contributing the architectural thinking and credibility that turns prospects into customers.
• Strengthen the ML practice from the inside — through peer guidance, technical interviews, and contributions to accelerators, reference architectures, and thought leadership content.

What You Bring

The non-negotiables

• 10+ years in machine learning or AI, with a proven track record of leading client-facing engagements in a consulting or advisory capacity.
• Deep, current knowledge of the AWS ML and GenAI ecosystem, with the ability to make and defend architectural decisions across the full ML lifecycle — from data and feature engineering through training, deployment, and monitoring.
• Deep expertise in at least two or three ML domains — whether traditional ML, computer vision, NLP, time series, or others — combined with the judgment to assess, architect, and advise across the broader ML landscape.
• Proven ability to architect and govern production ML systems end-to-end, translating MLOps, LLMOps, and broader AI operations complexity into standards and decisions that engineering teams can execute and executives can act on.
• Deep expertise across foundation model adaptation — fine-tuning (LoRA, QLoRA, PEFT), alignment (RLHF, DPO), inference optimization (quantization, vLLM), and distributed training (DeepSpeed, FSDP) — combined with RAG and agentic system design, including multi-agent architectures, event-driven workflows, MCP integration, and human-in-the-loop patterns on AWS. Technical authority to prescribe the right approach and set architectural standards that teams can execute against.
• Proven ability to operate independently in complex customer environments — navigating ambiguity, aligning stakeholders, and translating ML tradeoffs into business risk and value for both technical and executive audiences.

Strong differentiators

• AWS Certified Machine Learning – Specialty and/or AWS Certified Solutions Architect – Professional.
• Experience shaping practice-level standards, reference architectures, and reusable ML accelerators across multiple engagements.
• Exposure to varied industries and problem types in a consulting or client-facing context.
• Deep fluency in responsible AI practices — model evaluation, bias detection, fairness frameworks, and AI governance — applied in enterprise deployments.
• Hands-on experience designing and deploying SRE agents and AI-driven operations workflows in production — spanning automated incident detection, triage, and remediation — with the ability to integrate across observability platforms and translate AI operations outcomes into measurable business value.

Technical Stack

Our practice spans a broad range of ML domains. Candidates are expected to prescribe — not just recognize — with the judgment to maximize what AWS makes possible and the experience to know how open-source tooling strengthens it.

ML Domains: Classical ML, Computer Vision, NLP, Generative AI & LLMs, AI Agents & Autonomous Systems, Intelligent Document Processing, Video Understanding, Speech & Audio, Time Series & Forecasting, Recommender Systems, Graph ML, Reinforcement Learning, Multimodal AI

AWS ML Platform: SageMaker, SageMaker Pipelines, SageMaker Feature Store, SageMaker Model Registry, SageMaker Clarify, Bedrock (Agents, Knowledge Bases, Guardrails, AgentCore, Model Evaluation)

Multi-provider LLM: Bedrock, Anthropic API, OpenAI API, Google Gemini API, Azure OpenAI — with the judgment to reason across provider tradeoffs in enterprise contexts

AWS AI Services: Rekognition, Comprehend, Transcribe, Textract, Translate, Personalize, Neptune, Kinesis Video Streams, Polly

Data Platform: Apache Spark / PySpark, Apache Kafka, Amazon Kinesis, Apache Iceberg, Delta Lake, Apache Hudi, AWS Glue

Vector Databases: Pinecone, pgvector, Amazon OpenSearch (vector), Weaviate

Frameworks: PyTorch, TensorFlow, JAX, Scikit-learn, XGBoost, HuggingFace (Transformers, PEFT, TRL), LangChain, LlamaIndex, DSPy, Ollama

MLOps & Governance: MLflow, W&B, Airflow / MWAA (data orchestration), Dagster (asset-based pipelines), Kubeflow Pipelines, CI/CD, IaC (CloudFormation, CDK, Terraform), Docker, Kubernetes, ML Governance (lineage, data contracts, audit), Responsible AI / Bias & Fairness

LLM Evaluation & Safety: RAGAS, LLM-as-judge patterns, DeepEval, NeMo Guardrails, Constitutional AI patterns, structured output validation

Inference & Optimization: Triton, vLLM, SGLang, Trainium, Inferentia, Quantization (GPTQ, AWQ, bitsandbytes), SageMaker Neo

Benefits