Solution Engineering & DeliveryTranslate high-level designs into clear component contracts, APIs, and service boundaries. Implement LLM integrations, RAG pipelines, agents, tool/function calling, and prompt strategies. Own feature delivery for sprints/releases; maintain high code quality and documentation.Modeling & EvaluationFine-tune models when needed; design evaluation harnesses and metrics. Build A/B testing setups; track accuracy, latency, robustness, and task success rates. Conduct error analysis; iterate using feedback efficacy loops and prompt refinement.Data & Retrieval EngineeringBuild ETL/ELT pipelines; curate datasets with metadata, lineage, and validation. Implement vector indexing (chunking, embeddings, reranking), tune chunk size & overlap. Enforce data governance: PII handling, redaction, consent, auditability.MLOps & Platform ReadinessContainerize workloads (Docker); orchestrate deployments (Kubernetes/Helm). Own CI/CD for ML: train ?
evaluate ? package ? deploy ? monitor ? rollback. Maintain model/agent registries, experiment tracking, and reproducible environments.Software Engineering & IntegrationBuild microservices and async inference paths; support batch/stream processing. Integrate with enterprise auth, observability, telemetry, and logging. Write unit/integration/e2e tests, performance benchmarks, and failure-injection tests.Observability, Reliability & PerformanceInstrument with metrics/logs/traces; define SLOs (latency, throughput, error rate).
Optimize inference: batching, caching (KV cache), quantization, token efficiency. Implement guardrails (safety filters, jailbreak detection), auto-evals and alerts.Security & ComplianceApply secure coding practices; manage secrets, encryption, and least privilege. Ensure compliance (data residency, consent, audit trails); respect IP policies. Enforce policy-based access and content safety in user-facing features.Collaboration & MentoringReview designs/PRs; coach L3 engineers on best practices. Coordinate with AI Architects, Data Engineers, QA, and Product.Education and Experience RequiredBachelor's or master's degree in computer science, engineering, data science, machine learning, artificial intelligence, or closely related quantitative discipline. Typically, 7-10 years' experience.Knowledge and SkillsLLMs & Agents: Prompt engineering, function/tool calling, orchestration frameworks, RAG. ML/DS: Evaluation metrics (precision/recall, BLEU/ROUGE where relevant), error analysis. Data/RAG: Embeddings, similarity (cosine/IP), chunking, rerankers, vector DB operations.
Backend: Python (FastAPI/Flask), microservices patterns. MLOps/Infra: Docker, Kubernetes, CI/CD, artifact management, GPU scheduling. Observability: Metrics/logging/tracing, dashboards, automated evaluation pipelines. Frameworks: PyTorch/TensorFlow, Hugging Face, LangChain/LlamaIndex. Data: Pandas, SQL/NoSQL, Parquet/Arrow, Kafka/queues. Vector DBs: FAISS, Milvus, pgvector, Pinecone, Weaviate. Ops: GitHub Actions/Azure DevOps, MLFlow/W&B