AI工程师总监

职位说明

Role Summary

Lead dentsu's applied AI engineering team for the Data Experience practice.

You will own the full stack of what we ship: web applications, the agentic harness underneath them, and model hosting and observability on Databricks.

You will manage a senior India

based team of AI developers and full

stack engineers, set the engineering standard, and stay hands

on in the code.

This is a builder

leader role: you design supervisor graphs, review pull requests, tune context and token budgets, and ship agentic products that run in production, not demos.

Key Responsibilities

Lead and develop a senior India-based AI engineering team — two Directors and three Senior Managers (AI developers), two full-stack engineers, and a front-end developer; set priorities, run delivery across time zones, and grow the bench

Build full-stack web applications end to end — Python services, APIs, and modern front ends that put agentic capability in front of internal teams and clients

Engineer the agentic harness — the runtime layer that manages tool calling, state, memory, retries, guardrails, human-in-the-loop checkpoints, and evaluation for every agent we ship

Design multi-agent systems in LangGraph — supervisor node structures, sub-agent routing, agent-to-agent (A2A) communication, and shared state across long-running workflows

Ship agentic features inside web apps — embedded assistants, tool-using workflows, and streaming UX that feel native to the product rather than bolted on

Own context engineering — what goes into the context window, when, and why: retrieval, summarization and compaction, tool-result shaping, and memory strategy

Drive token management and optimization — prompt caching, model routing by task, context budgets, and per-feature cost and latency tracking

Host and operate models and agents on Databricks — Model Serving, Unity Catalog governance, and MLflow 3 tracing and evaluation for end-to-end observability

Set the AI-native development standard — Claude Code, Cursor, and Codex as daily tools; codify rules, prompts, and review practices so the team ships faster without sacrificing quality

Run engineering hygiene through GitHub — branching strategy, code review, CI/CD with GitHub Actions, automated testing, and environment promotion from dev to production

Partner with product, data engineering, and client teams — translate business problems into agent designs and explain trade-offs to non-technical stakeholders

Required Qualifications

8+ years in software engineering, including 2+ years building LLM-powered or agentic systems in production, with experience leading senior engineers

Strong hands-on experience with the Claude Agent SDK or the OpenAI Agents SDK (or equivalent) — building and shipping agents, not just calling a completions endpoint

Proven experience engineering agentic harnesses — tool orchestration, state management, error recovery, guardrails, and evaluation

Deep context engineering skills, including token management and optimization and prompt caching strategies

Expert Python and strong LangGraph experience, including supervisor and multi-agent patterns and agent-to-agent communication

Comfortable building full-stack web applications — Python backends (FastAPI or similar), REST and streaming APIs, and a modern JavaScript front end (React / Next.js or similar)

Experience deploying models and agents on Databricks with MLflow, including MLflow 3 tracing and evaluation for observability

Fluent with AI coding tools — Claude Code, Cursor, or Codex as part of your daily workflow, and able to teach others to use them well

Solid DevOps fundamentals on GitHub — CI/CD pipelines, GitHub Actions, code review, testing, and secrets and environment management

Track record managing distributed or offshore engineering teams across time zones

Preferred Qualifications

Data engineering background — Spark / PySpark, lakehouse architecture, SQL, and working with governed gold data products

Experience building Model Context Protocol (MCP) servers or implementing the A2A protocol

LLM evaluation practice — offline eval sets, LLM-as-judge, and regression testing for prompts and agents

Experience with Databricks Apps and Lakebase (Postgres)

Experience in media, advertising technology, or marketing analytics environments

Additional Information

The annual salary range for this position is $94,000 - $170,000.

Placement within the salary range is based on a variety of factors, including relevant experience, knowledge, skills, and other factors permitted by law.

#LI-Hybrid #LI-MS3

Benefits available with this position include

Medical, vision, and dental insurance

Life insurance

Short-term and long-term disability insurance

Flexible paid time off

At least 15 paid holidays per year

Paid sick and safe leave, and

Paid parental leave.

Dentsu also complies with applicable state and local laws regarding employee leave benefits, including, but not limited to providing time off pursuant to the Colorado Healthy Families and Workplaces Act, in accordance with its plans and policies.

For further details regarding Dentsu benefits, please visit www.

dentsubenefitsplus.

Location

New York - 150 E 42nd

Dentsu Media

Time Type

Full time

Contract Type

Permanent

Dentsu

is committed to

providing

equal employment opportunities to all applicants and employees. We do this without regard to race

national origin

is committed to providing reasonable accommodation to, among others, individuals with disabilities and disabled veterans. If you need an accommodation because of a disability to search and apply for a career opportunity with us, please send an e-mail to

ApplicantAccommodations@dentsu.com

by clicking on the link to let

us know

the nature of your accommodation request and your contact information. We are here to support you.