LLM Application Engineer
Build production features on top of foundation models: retrieval, tool use, evaluation and cost control.
- Typical band
- USD 4,000 – 8,500 /mo
- Markets
- Mexico, Colombia, Argentina, Brazil
- Mode
- Remote across LATAM
Careers, AI tracks
SILA places AI engineering, data, evaluation and AI-operations professionals with US and global companies, employed compliantly in your own country, paid on time, in your currency or in USD. Twenty one tracks, all of Latin America, one recruiter who answers you.
Bands are the monthly ranges we typically see for these tracks across the region. The final offer depends on the client, the market and your level, and we tell you the number before you interview.
Build production features on top of foundation models: retrieval, tool use, evaluation and cost control.
Train, fine-tune and ship models with real serving constraints and measurable business outcomes.
Detection, OCR, video analytics and edge inference for industrial, retail and logistics use cases.
Spanish and Portuguese language systems: classification, extraction, summarization and speech pipelines.
Translate business problems into model, data and integration architecture with a defensible cost model.
CI/CD for models, feature stores, monitoring, drift detection and inference cost governance.
Internal tooling, GPU scheduling, orchestration and developer experience for AI teams.
Ingestion, transformation and warehouse design so models train on data that can be trusted.
Modeled metrics, semantic layers and reporting that survive audit and executive scrutiny.
Experiment design, forecasting and causal analysis tied to a commercial decision, not a dashboard.
Prompt systems, context assembly, regression suites and measurable output quality.
Human review of model output, hallucination and safety checks, escalation and scoring rubrics.
Bilingual labelling for text, image, audio and video datasets with documented quality gates.
Spanish and Portuguese preference data, comparison ranking and reasoning-trace authoring.
Adversarial testing, jailbreak discovery and safety documentation for regulated deployments.
Tier 1 and 2 support for AI products, with feedback loops back into evaluation datasets.
Bilingual assistant flows, escalation logic and tone standards for customer-facing AI.
Own the roadmap where model capability, cost and user trust intersect.
Agentic workflows, integrations and back-office automation with human-in-the-loop controls.
Model inventories, data-protection mapping and audit evidence for enterprise AI programs.
Run nearshore AI pods end to end: scope, staffing, delivery cadence and client reporting.