Technical Architect - ML

Quantiphi; Princeton (NJ); On-site; USD 180,000 - 240,000; Full time

Job summary

Quantiphi is seeking an experienced Architect Machine Learning Engineer in New Jersey (Onsite) to architect, build, and deploy production-grade agentic AI systems and multi-agent workflows from the ground up.

You will design autonomous AI systems that collaborate, reason, and execute complex tasks with minimal human intervention, deploying on major cloud platforms and enforcing MLOps practices for reliability and scalability.

Qualifications

• 6–8 years hands-on ML/AI production experience
• Proven ability to deploy production-grade agentic AI systems
• Experience with Langraph or CrewAI is a plus

Responsibilities

• Architect and build end-to-end multi-agent systems from scratch.
• Develop agent tools, skills, and orchestration layers.
• Implement memory and context engineering for stateful agents.
• Deploy low-latency agentic systems on AWS/GCP/Azure.
• Apply MLOps practices: CI/CD, monitoring, and cost optimization.
• Integrate LLMs as core reasoning engines with RAG/PEFT.

Skills

Python programming

ML engineering

Multi-agent systems

Cloud platforms

CI/CD

Observability

Tools

Pinecone

Weaviate

ChromaDB

Langraph

CrewAI

LangSmith

Terraform

Docker

Kubernetes

Job description

**Role:**Architect Machine Learning Engineer **Experience**: 8–12 Years **Location**: New Jersey (Onsite) **Job Summary**: We are seeking an experienced Architect Machine Learning Engineer to architect, build, and deploy production-grade agentic AI systems and multi-agent workflows from the ground up. The ideal candidate will have deep expertise in designing autonomous AI systems that can collaborate, reason, and execute complex tasks with minimal human intervention. You will be responsible for creating scalable, robust agentic workflows using cutting-edge frameworks like CrewAI/Langraph, while ensuring enterprise-grade deployment on major cloud platforms. **Roles & Responsibilities**:● Architect & Build Agentic Systems: Design and develop end-to-end multi-agent systems from scratch. You will create the foundational agent harnesses, define communication protocols, and build orchestration layers using frameworks like CrewAI, Langgraph, and AutoGen. Architectural decisions to ensure:○ Hierarchical and collaborative multi-agent structures with well-defined agent roles, responsibilities, and communication protocols○ Dynamic task decomposition, sophisticated tool integration, planning mechanisms (ReAct), and self-correction loops○ Develop state management systems and memory mechanisms for persistent agent interactions● Engineer Advanced Agent Capabilities: Develop custom agent-tools and define specialized agent-skills that empower agents to perform complex, domain-specific tasks.● Pioneer Context Engineering: Implement advanced context engineering and memory systems to ensure agents maintain state, learn from interactions, and make informed decisions in dynamic environments.● Deploy Production-Grade Solutions: Own the deployment, scaling, and maintenance of robust, low-latency agentic systems on major cloud platforms (GCP, AWS, or Azure). You will implement best-in-class MLOps practices for monitoring, continuous integration/continuous deployment (CI/CD), and system reliability.● Integrate and Optimize LLMs: Integrate LLMs to serve as the core reasoning engines for autonomous agents. You will apply advanced techniques like RAG and PEFT to optimize performance.● Create and maintain comprehensive tool libraries for agents including API integrations, database queries, and external service connections● Design and implement RAG systems using vector databases (Pinecone, Weaviate, ChromaDB)● Develop custom tools and plugins that enable agents to interact with various enterprise systems and APIs● Ensure tool reliability, error handling, and seamless integration within agentic workflows● Implement comprehensive monitoring and tracing systems for agent behavior, performance, cost optimization, and latency analysis ● Design novel evaluation frameworks to assess multi-step agentic task success, reliability, and accuracy● Utilize advanced observability tools (LangSmith, Arize AI, or custom solutions) to trace agent decision making processes● Establish metrics and KPIs for measuring agentic system performance in production environments **Required Skills & Qualifications**: **Experience**:● 6-8 years of hands on experience in machine learning and AI engineering with proven track record of taking ML systems to production● Demonstrated expertise in building multi-agent systems and agentic workflows, preferably with Langraph/CrewAI Technical Skills - Must Have:● Programming & ML: Expert-level Python proficiency with ML frameworks (TensorFlow, PyTorch, Transformers). Experience with FastAPI, async programming, and microservices architecture● Data & Vector Systems: Hands-on experience with vector databases (Pinecone, Weaviate, ChromaDB) and building scalable RAG systems● Monitoring & Observability: Experience with LLM application monitoring tools (LangSmith, Weights & Biases, custom telemetry solutions)● Proven ability to architect and implement complex AI systems from scratch in production environments● Cloud Platform Expertise: Production-level experience with at least one major cloud platform (AWS, GCP, or Azure), including: ○ Compute services (EC2, GCE, Azure VMs) ○ Serverless functions (Lambda, Cloud Functions, Azure Functions) ○ Container orchestration (EKS, GKE, AKS) ○ Managed AI/ML services (SageMaker, Vertex AI, Azure ML)● Production & DevOps: Strong skills in Infrastructure as Code (Terraform, CloudFormation), CI/CD pipelines (GitHub Actions, Jenkins), and containerization (Docker, Kubernetes) **Technical Skills** - Good to have: ● Experience with prompt engineering techniques, fine-tuning SLMs (PEFT, SFT, RLHF), and model optimization● Knowledge of distributed systems, message queues, and event-driven architectures for agent coordination● Familiarity with SDLC best practices, version control (Git), and agile development methodologies● Experience with tool-calling agents, multi-step workflows, and stateful orchestration (e.g. graphs, planners, routers).● Hands-on evals for agents: trajectory / tool-use checks, golden traces, LLM-as-judge with fixed rubrics, regression suites.● Online evals, drift thinking, and clear quality gates before or after deploy (thresholds, alerts, rollback criteria).● Safety and abuse: prompt injection via tools, untrusted retrieval, PII handling in prompts and logs, allowlists and guardrails.● Cost and latency discipline: budgets per run, timeouts, caps on turns and tool calls.● Model lifecycle: routing / gateway patterns, version pinning, fallbacks, and which model for which step.● Memory and state: what is persisted, retention, redaction, and what must never be stored*If you like wild growth and working with happy, enthusiastic over-achievers, you'll enjoy your career with us**!*