Gusto, Inc. · , Denver, CO - Hybrid · New York, NY - Hybrid
$132k–$200k a year · Hybrid · Lead & above
Medium competition (est.)
Filed H-1B · FY25Filed 11 H-1B LCAs in FY2025
Posted 5 days ago
Contribute to agent frameworks and developer harnesses that support tool calling, structured outputs, reusable skills, actor-based workflows, human review, and policy-controlled execution.
Implement services that dynamically inject relevant knowledge into agent workflows based on investigation type, entity context, workflow state, customer policy, prior findings, and current task requirements.
Build domain-grounding capabilities that help agents distinguish authoritative facts, customer-defined policies, derived signals, hypotheses, and unverified observations.
Develop knowledge-ingestion and enrichment pipelines that normalize, classify, chunk, index, version, and refresh structured and unstructured data.
Implement retrieval workflows using combinations of semantic search, lexical search, metadata filtering, hybrid retrieval, reranking, entity-aware retrieval, and temporal filtering.
Contribute to knowledge-graph systems that represent entities, relationships, events, evidence, provenance, and risk signals.
Support graph traversal, multi-hop retrieval, graph-enhanced context assembly, and integration between graph, vector, search, document, and API-based data sources.
Implement context-selection logic that balances relevance, freshness, source authority, latency, tenant permissions, and model context limits.
Add safeguards against stale, contradictory, duplicated, irrelevant, poisoned, or unauthorized information entering agent contexts.
Build state and memory capabilities for agents, including short-term task state, investigation history, entity context, retrieval state, and provenance.
Integrate agents with internal APIs, identity and risk data sources, case-management systems, watchlist systems, model endpoints, webhooks, and event streams.
Develop secure connector and tool-execution patterns, including authentication, authorization, credential management, tenant isolation, audit logging, and sandboxed execution where appropriate.
Integrate machine-learning models used for classification, ranking, extraction, entity resolution, summarization, and other investigation-related capabilities.
Partner with machine-learning and data-science teammates on feature definitions, training data preparation, labeling strategies, dataset quality, model interfaces, and deployment workflows.
Contribute to model training, fine-tuning, and adaptation workflows, including dataset preparation, experiment tracking, evaluation, versioning, deployment, and rollback.
Implement model-serving and model-routing integrations that select models based on task requirements, latency, cost, quality, context, and data boundaries.
Improve system performance through caching, batching, parallel execution, retrieval optimization, context reduction, model selection, prompt optimization, and efficient inference configuration.
Build evaluation and regression tests for retrieval quality, grounding, citation correctness, model behavior, tool selection, workflow completion, and end-to-end investigation outcomes.
Create test cases covering domain-specific scenarios, ambiguous cases, adversarial inputs, policy variations, knowledge freshness, retrieval failures, and production incidents.
Sponsorship data shows employers' past filings in public US government records (DOL, USCIS, E-Verify) and is not a promise that any job will be sponsored. Immigration rules change. Verify your situation with your DSO or an immigration attorney.