Overview
Own the technical design and delivery of complex AI applications for business and customer needs, as a hands-on senior individual contributor who leads architectural decisions independently.
What you'll do
- Translate ambiguous requirements into robust system designs, service boundaries, data contracts, and implementation plans.
- Lead build vs. buy decisions and document trade-offs across model quality, latency, throughput, reliability, security, and cost.
- Design retrieval-augmented generation pipelines using vector databases, sparse and dense retrieval, hybrid search, metadata filtering, and reranking.
- Define document parsing, chunking, embedding, and indexing strategies.
- Determine when model adaptation is warranted over prompting or retrieval; curate training and validation datasets, implement supervised fine-tuning and parameter-efficient techniques such as LoRA, and benchmark against baseline models for task accuracy, safety, latency, and cost.
- Build critical Python/TypeScript services, APIs, and asynchronous workflows.
- Partner with Data Engineering on ingestion and indexing pipelines, including normalization of customer data, schema validation, incremental updates, lineage, and recovery.
- Develop stateful agentic workflows with task decomposition, planning, multi-step tool use, dynamic tool selection, structured outputs, memory management, and multi-agent orchestration where justified.
- Build tool and MCP integrations with validated inputs, scoped permissions, and human approval for sensitive actions.
- Evaluate retrieval and agent quality against defined accuracy and grounding metrics.
- Implement defenses against prompt injection, sensitive-data leakage, unauthorized tool execution, and retrieval access violations; contribute threat models, adversarial tests, and evidence for governance reviews.
- Translate centrally defined policies into application controls for identity, tenant isolation, data residency, retention, and auditability.
- Work with Engineering and forward-deployed teams to implement cloud, hybrid, and customer-controlled deployments.
- Define evaluation datasets and measure retrieval quality, groundedness, task completion, tool correctness, latency, and cost, combining deterministic checks, calibrated LLM-as-judge, and human review.
- Partner with LLMOps on CI/CD release gates, prompt and model versioning, tracing, and safe rollouts and rollbacks.
- Set and enforce quality standards across the team, with accountability for code quality, maintainability, and production reliability.
- Lead architecture and code reviews, mentor AI Specialists, and drive root-cause analysis across model, retrieval, application, and data layers.
- Improve developer productivity through reusable libraries, reference architectures, tooling, and runbooks.
What you'll need
- 6+ years in applied AI, machine learning, or software development, including architecture and production ownership of complex AI systems.
- Experience applying enterprise AI controls, including scoped tool authorization, tenant isolation, PII handling, prompt-injection defenses, audit logging, and human approval for sensitive actions.
- Experience designing enterprise semantic layers or context graphs that connect business entities, relationships, and definitions across systems.
- Strong experience with cloud deployments on GCP/AWS, including serverless technologies, CI/CD pipelines using GitHub, Docker and Kubernetes, and model platforms such as Vertex AI, Bedrock, etc.
- Production-grade development using Python, TypeScript, and SQL, with typed interfaces, APIs, asynchronous processing, and automated testing.
- Build enterprise agentic systems using LangGraph, Mastra, or equivalent frameworks, including planning, multi-agent orchestration, state and memory management, checkpointing, and failure recovery.
- Applied knowledge of LLM inference and context engineering, including structured outputs, embeddings, hybrid retrieval, reranking, and evaluation-driven model selection or fine-tuning.
- Practical work with Databricks, Spark, and Postgres, spanning analytical and transactional workloads, batch and real-time pipelines, and normalization of messy customer data.
- AI evaluation using Braintrust or equivalent, including golden datasets, custom scorers, regression testing, and calibrated LLM-as-judge to measure task success, tool correctness, safety, and cost.
- Experience with enterprise AI observability layers, including agent traces, model and tool spans, prompt versions, error attribution, latency, token usage, and cost tracking.
Nice to have
- Familiarity with centralized LLM gateways such as LiteLLM, covering multi-provider routing, fallbacks, rate limits, token budgets, caching, and access policies.
Details
Read the full description and apply on the company’s own careers page.