AI Orchestration Engineer

ReAct, plan-and-execute, reflexion, and reflexion-based agent topologies.

Message passing, shared memory blackboards, and inter-agent protocol design.

Defining tool schemas, binding tools to models, and handling structured outputs.

Few-shot, chain-of-thought, and instruction tuning for reliable model behavior.

Chunking strategies, hybrid retrieval, reranking, and context window management.

Embedding model selection, indexing (HNSW, IVF), and metadata filtering.

LLM-as-judge, human eval panels, regression suites, and drift detection.

Schema validation, safety classifiers, and fallback response strategies.

Short-term context, episodic memory, and long-term knowledge consolidation.

DAG-based pipelines, conditional branching, and human-in-the-loop checkpoints.

Cost-aware routing, small-to-large cascades, and fallback model strategies.

Token streaming, partial JSON parsing, and progressive UI rendering.

Tracing spans, token cost tracking, and latency profiling across agent calls.

Caching, prompt compression, and batch inference for production cost control.

Blue-green deploys, shadow traffic, and rollback for agent workloads.