Machine Learning Engineer – NC or VA preferred OurclientisseekinganexperiencedMachineLearningEngineertoleadthedesignandimplementationofproduction-readyAIsystemsfocusedondocumentintelligence,agenticworkflows,andsecuredeployment.Thisrolewillinvolvebuildingsystemsthattransformunstructuredtechnicalcontentintostructuredandactionabledatamodelssuchasmodulesandproceduralsteps.YouwillberesponsiblefordevelopingAIagentscapableofmulti-stepreasoning,integratingrule-based systems with large language models, and orchestrating decision flows that call the appropriate models,orfallbacklogic.Youshouldbecomfortabledesigningpipelinesthatoperatewithoutaccesstothird-partyAPIsandinsteadrunfullyon-premiseorinisolatedenvironments.Youwillcollaboratecloselywithbackendanddataengineerstotranslatecomplexandoftenambiguousproductrequirementsintoworkingprototypesandscalableworkflows.Thesesolutionsmustmeethighstandardsofperformance,reliability,andmaintainability.Thisopportunityisadirect hire opportunitywithasalaryrangestartingat$150,000whichincludesasolidbenefitspackage.CandidatescanworkremotelybutpreferlocaltoNCorVA(however,willacceptNY,DCorsouthofDC candidates). US Citizenshipisa mus tsopleasedonotapplyifyouareonanytypeofvisanoworinthefuture. Key Responsibilities Build and deploy retrieval-augmented generation (RAG) pipelines that ground general-purpose models in proprietary documents Design agentic workflows using LangChain, LlamaIndex, or similar tools to support multi-step reasoning and tool orchestration Engineer hybrid inference strategies that combine lightweight task-specific models, rules, and LLM components based on context and confidence Fine-tune LLMs using domain-specific data when appropriate, while using prompt engineering and safety guardrails to handle ambiguous user input Translate high-level product goals (e.g., "turn this folder of technical documents into a searchable training module") into structured experiment plans Develop data ingestion pipelines to handle scanned PDFs, forms, structured logs, and other industrial data types Integrate ML pipelines with backend systems (Django/PostgreSQL) and Snowflake to support end-to-end deployment Create observability layers including model latency, drift detection, and human-in-the-loop feedback systems Ensure all ML systems meet performance and security constraints for deployment in air-gapped and field environments Minimum Qualifications 5+ years of experience in machine learning, with a strong track record of owning and delivering ML systems end to end Hands-on experience with LLM agents or similar systems using LangChain, LlamaIndex, or custom-built orchestration logic Demonstrated experience building production-ready RAG pipelines and/or embedding-based semantic search systems Familiarity with prompt design strategies including few-shot, chain-of-thought, and fallback logic under safety constraints Strong grounding in ML fundamentals, including supervised learning, tree-based models, and linear models Proficiency in Python and libraries like PyTorch, Hugging Face Transformers, and Scikit-learn Experience with feature engineering for sparse or noisy data Proficient in deploying containerized ML systems (Docker/Kubernetes) with CI/CD pipelines and basic observability practices Strong experience with ETL tools and orchestration frameworks such as Airflow, Dagster, or Luigi Comfort working with data in less-than-perfect form, and a mindset that 80% of the job is cleaning up someone else’s CSV U.S. Citizenship and ability to obtain a security clearance (Secret or TS/SCI eligible) Preferred Qualifications Experience deploying models in air-gapped or disconnected environments Familiarity with OCR and layout-aware document understanding techniques Experience using or building vector databases and custom embedding strategies Ability to balance model-driven and rule-based approaches when designing ML systems Exposure to aviation, defense, or industrial workflows where high assurance and compliance are critical Experience designing feedback loops and evaluation workflows to continuously improve model performance in production