AI Engineer- NLP
Velocitor Solutions3 months ago
Charlotte, NC, USAMid Level / Senior
Responsibilities
- Extend the guardrails, query reformulation, embedding-based routing, LangGraph agent, response formatting, and follow-up generation pipeline.
- Maintain and expand the VTrack API domain tool layer, including schemas, authorization, pagination, date handling, and error formatting.
- Support the production service by responding to incidents, investigating latency and quality regressions, and improving telemetry and runbooks.
- Improve RAG retrieval quality using PostgreSQL full-text search, pgvector, hybrid retrieval, and reranking where appropriate.
- Build evaluation gates for model and prompt changes using representative and adversarial datasets, shadow traffic, canary rollout, and automated rollback.
- Reduce LLM cost and latency through token and cost telemetry, prompt and context trimming, caching, model tiering, and removal of redundant LLM stages.
- Strengthen tenant-scoped authorization, prompt-injection defenses, and security boundaries across prompts, retrieval, tools, authorization, and outputs.
- Extend tiered testing across commit, pull request, nightly, and release gates.
Requirements
- Three or more years of experience building and supporting production backend services, including hands-on delivery of at least one LLM-backed feature used by real users.
- Strong Python skills, including asynchronous programming, type-driven design, and work in a strict mypy codebase.
- Experience with an LLM orchestration or agent framework such as LangChain, LangGraph, or an equivalent, including tool and function calling.
- Solid PostgreSQL experience covering schema design, query performance, migrations, and connection-pool behavior under load.
- Production service support experience, including telemetry-based diagnosis, blast-radius assessment, and rollback decisions.
- Understanding of when to use autonomous agents versus deterministic workflows, especially for data-modifying or compliance-sensitive operations.
- Understanding of AI-system security boundaries, untrusted model and retrieved content, external authorization enforcement, and tenant-scoped data access.
- Ability to debug across service boundaries using traces, latency metrics, and correlation IDs.
- Familiarity with retries, jittered backoff, circuit breakers, and concurrency limits for rate-limited upstream providers.
- Testing experience beyond unit tests, including external API contract tests and evaluation of nondeterministic components.
- Prior production-system handover experience is preferred.
- Azure experience, especially Azure Container Apps, Azure OpenAI deployments and quota management, and Key Vault, is preferred.
- Terraform or OpenTofu, Azure DevOps Pipelines, vector search and RAG at scale, LLM-as-judge evaluation, and NeMo Guardrails or equivalent safety-layer experience are preferred.
- Modern React and TypeScript experience and data-retention or privacy-engineering experience are preferred.
Tech Stack
Categories
About Velocitor Solutions
Velocitor Solutions builds enterprise mobile and fleet management software and services for transportation, logistics, field service, and route accounting teams. Its V Track platform combines GPS tracking, AI-enabled video telematics, and real-time data, delivered with managed services such as hardware procurement and hosting. Founded in 2000 and headquartered in Charlotte, NC, the company reports managing over 200,000 mobile assets for Fortune 500 clients across North America.