Acceleration Center - Agentic AI & DevOps Engineer - Senior Associate
pwc · Mexico City ·
- Seniority
- Senior
- Category
- Devops
- Experience
- 2+ years
- Company size
- 1000+
pwc · Mexico City ·
PwC México · Mexico
BLEN · United States
Ford Motor Company · Dearborn, MI, United States
Greenlight Consulting · Ontario, Canada
Build and deploy LLM-powered agentic systems, RAG pipelines, and MLOps tooling end-to-end, ensuring scalable, observable, and reliable AI solutions in production.
Industry/Sector
Not ApplicableSpecialism
Data, Analytics & AIManagement Level
Senior AssociateJob Description & Summary
The Opportunity
Join our Acceleration Center Mexico and help shape the future of business for our diverse client portfolio across geographies and jurisdictions. You’ll work at the heart of global teams across Advisory, Assurance, Tax and Business Services—solving real client challenges through connected collaboration. We’ll help you grow your skills so you can go further. With hands-on learning, cutting-edge tools and an inclusive culture, this is your opportunity to do inspiring work that makes a difference—every day.
As an Acceleration Center - Agentic AI Engineer - Senior Associate, you will play a pivotal role in transforming raw data into actionable insights, enabling informed decision-making and driving business growth. Within our Internal Firm Services practice, you will leverage Artificial Intelligence to solve business problems through Agentic Systems, RAG, LLMs, while ensuring solution scalability and correct implementation.
As a Senior Associate, you will focus on building meaningful client connections and learning how to manage and inspire others. You will navigate increasingly complex situations, growing your personal brand and deepening your technical skills. You are expected to anticipate the needs of your teams and clients, delivering quality work while embracing increased ambiguity. This role requires you to respond effectively to diverse perspectives, use a broad range of tools and methodologies to generate new ideas, and uphold professional and technical standards.
In this role at PwC Acceleration Center Mexico, you'll ship a system. This seat owns the model layer end-to-end: prompt and skill design, evals, safety, and the MLOps that keeps it cheap, fast, traced, and reproducible in production.
Responsibilities
- Designing and deploy LLM-powered automations at scale — agents, skills, pipelines — from prototype through production
- Building eval frameworks: offline benchmarks, adversarial red-teaming, regression suites, and production feedback loops that catch drift before users do
- Owning observability end-to-end: distributed traces, latency histograms, cost attribution, token throughput, error budgets, and dashboards that surface signal over noise
- Defininge and enforce SLOs for model-backed services; own the alerting stack and incident response playbooks
- Debuging production failures across the full stack — from Kubernetes pod restarts to model hallucinations to prompt injection — and write postmortems that actually prevent recurrence
- Operatinge AI-natively: use Claude Code, Codex, GitHub Copilot, Cursor, and similar tools as daily infrastructure, not novelty — and hold the output to the same standard as hand-written code
- Instrumenting and monitoring LLM behavior in production: token usage, response quality, failure modes, latency SLAs Architect CI/CD for model and prompt changes: versioning, automated eval gates, rollback strategies
What You Must Have
- At least a Bachelor's degree
- At least 2 years of experience
- Oral and written proficiency in English required
What Sets You Apart
- Core engineering: Python fluency, FastAPI, async patterns, Pydantic, Typescript, SQL/Redis for state and caching, K8s + Azure DevOps CI/CD
- ML/LLM Depth: strong fundamentals (transformers, finetuning, RAG, agent frameworks); hands-on with Claude/OpenAI SDKs/Langchain/Langgraph/Langfuse; building evaluations: prompt tests, LLM-as-a-judge pipelines, human feedback integration.
- MLOps/SRE: shipped LLM-backed automations to productions with real SLAs and real users, reproducible experiments (tracked, versioned, auditable), SLO definition, error tracking, alerting design, monitoring pipelines for model degradation, prompt drift and cost anomalies.
- AI Native Fluency: daily user of AI coding assistants (Claude Code, Codex, Copilot, Cursor), understandins model limitations, failure modes and when to distrust output, can write prompts and agent instructions with rigor, has opinions on eval design, context management and tool-use patterns.
Travel Requirements
Up to 20%Job Posting End Date
octubre 27, 2026Veeam · San Jose, CA, USA