Solution Architect

The AI Solutions Architect tasks include:

• Architect & Design end to end Generative AI solutions by translating business requirements into scalable, secure, and cost effective technical architectures

• Own solution design across the full lifecycle, including problem definition, architectural choices, data strategy, model selection, integration, deployment, and operations

• Define and document solution architectures using architecture diagrams, sequence flows, deployment topologies, and technical design documents

• Select appropriate GenAI approaches (RAG, fine tuning, prompt engineering, agentic workflows, routing, multimodal) based on accuracy, latency, cost, governance, and data constraints

• Architect Retrieval Augmented Generation (RAG) solutions, including document ingestion, chunking strategies, embedding models, hybrid search, re ranking, caching, and freshness management

• Design and implement agentic systems using tool/function calling, planner executor and multi agent patterns, with reliability, observability, and failure handling

• Define prompt architecture patterns such as ReAct, Chain of Thought alternatives, structured output prompting, skill routing, and template versioning

• Evaluate and integrate LLMs from both closed providers and open weight models, including model routing, fallback logic, and token cost optimization

• Design fine tuning strategies (LoRA/QLoRA, PEFT, adapters) when prompting or RAG alone is insufficient, and define evaluation criteria for fine tuned models

• Architect low latency, high throughput serving patterns using batching, caching, speculative decoding, quantization, and GPU/CPU routing

• Design APIs, SDKs, and reusable platform components to enable consistent AI adoption across multiple teams and applications

• Integrate GenAI solutions with enterprise systems including identity, authorization, data platforms, event streams, and front end applications

• Define LLMOps practices covering prompt versioning, dataset management, experiment tracking, evaluation pipelines, regression testing, and observability

• Establish monitoring and tracing for quality, hallucinations, safety violations, latency, throughput, and cost across AI workflows

• Design safety, governance, and guardrail mechanisms including PII redaction, content filtering, prompt isolation, jailbreak defense, and audit logging

• Conduct feasibility and risk assessments covering data readiness, compliance, security exposure, vendor lock in, performance, and operational risk

• Lead technical discovery workshops and architecture reviews with product, engineering, data, and security stakeholders

• Mentor engineers and teams on GenAI patterns, solution design tradeoffs, and architectural best practices

• Evaluate tools, frameworks, and vendors, and drive build vs buy decisions based on technical and business constraints

• Maintain reference architectures, design standards, architectural decision records (ADRs), and reusable blueprints

• Drive continuous improvement of AI solution quality, reliability, scalability, and cost efficiency in production environments.

Preferred Technical Skills

• Person should have excellent design experience with designing in-house systems (connecting with different ERP systems, connecting the data with other systems, designing data model)- minimum 10 years (5 years solution architect, 4-5 years Machine learning & AI experience)

• GenAI frameworks: LangChain, LangGraph, AutoGen, Microsoft Agent Framework.

• Model providers: Azure OpenAI, Anthropic, Google, AWS Bedrock, Snowflake Cortex AI, AgentCore

• Retrieval & Vector DBs: OpenSearch, Pinecone, pgvector, Weaviate, GraphDBs (Neo4j, Neptune)

• Serving Stack: vLLM, TGI, Ray Serve

• LLMOps: Dataiku, MLflow, LangSmith, Weights & Biases

• Safety: Azure AI Content Safety, Guardrails, NeMo Guardrails

• Agents: LangGraph, AutoGen, CrewAI

The preference is to have someone close to Princeton for 2X per week for whiteboarding sessions