Christiam Ipanaque | AI Engineer & Instructor | Artificial Intelligence Institute

Christiam Ipanaque - AI Engineer and Instructor at the Artificial Intelligence Institute

Christiam Ipanaque is an Artificial Intelligence Engineer based in Seattle, Washington. He is the founder of Northbridge Engineering, a firm that builds production AI systems for businesses: from customer support automation to lead qualification engines and document processing pipelines. His work spans the full stack of modern AI engineering: LLM API integration, vector search and embedding pipelines, LangChain and LangGraph orchestration, agentic system design, cloud deployment on AWS, and the observability and evaluation infrastructure that keeps production systems reliable.

Before founding Northbridge Engineering, Christiam spent years building full-stack and backend systems, developing the software engineering discipline that underpins his approach to AI. He has worked alongside engineering teams at technology companies solving production problems that only appear when AI systems meet real users, real data, and real failure modes. He brings this production mindset to every course he teaches: every technique he demonstrates, every architecture decision he walks through, and every failure mode he warns about comes from a system he personally built, operated, or repaired in a production environment.

Christiam holds deep technical expertise spanning the modern AI stack: LLM APIs and prompting, embeddings and vector search, RAG pipeline architecture, LangChain and LangGraph state machines, agentic system design, model fine-tuning with LoRA, and cloud deployment on AWS with CI/CD, observability, and cost controls. He is equally focused on the engineering practices that make AI systems reliable in production: rate limiting, semantic caching, automated evaluation, and incident response.

GitHub LinkedIn Northbridge Engineering

Production Experience

Scenarios Christiam Has Solved in Production

Every engineer accumulates war stories. Christiam's are different: they are documented, measurable, and mapped directly to the six format families in the Institute's Scenario-Based Assessment standard. Each scenario below is a real situation he faced, not a hypothetical.

F2: Diagnosing a Degraded RAG Pipeline in Production

Christiam was called in when a deployed RAG system's retrieval quality dropped 40% overnight. Observability was partially broken, the regression was intermittent, and a stakeholder was pressing for a premature rollback. He repaired enough of the measurement to reason safely, isolated the cause to a drift in embedding distribution after a model provider deployment, implemented a fix, and communicated the timeline honestly. This scenario exercises D4 (Diagnosis & Improvement) under adverse conditions, generating D6 (Professional Judgment) evidence under pressure.

F3: Choosing Between Latency, Cost, and Accuracy at Scale

Faced with a real-time LLM agent whose per-request cost was running 6x above projections, Christiam had to redesign the architecture mid-deployment without sacrificing the accuracy guarantees the client contract required. He introduced semantic caching, model tiering (smaller models for routine queries, full-scale LLM for complex ones), and structured output routing: reducing cost by 70% while maintaining response quality. This is D5 (Design Under Constraint) exercised under a live deadline.

F1: Building an Agent from a Two-Sentence Request

A client asked for "an AI that qualifies leads." No data, no schema, no success criteria. Christiam scoped the requirements through stakeholder interviews, built a LangGraph-based agent with iterative refinement and human-in-the-loop validation, and shipped a system that processed 5,800+ leads per month with a 47% increase in qualified conversions. This scenario samples D1 (Systems Construction) and D5 (Design Under Constraint) from an underspecified starting point.

F4: Building a Measurement System Where None Existed

A customer support AI had been deployed for months, but no one could say whether it was good. Christiam designed automated evaluation harnesses measuring answer correctness, retrieval precision, escalation rate, and a user satisfaction proxy. He established a baseline the team had never had, then iterated the system against that measurement. This is D3 (Evaluation & Measurement) applied to a production system that had been running blind.

F5: Catching a Data-Handling Flaw in a Peer's Design

During an architecture review, Christiam identified that a proposed agent system would exfiltrate personally identifiable information through its tool-calling path. He documented the flaw, proposed a guardrail layer with output filtering and input sanitization, and escalated the risk before the system reached production: an instance of D6 (Professional Judgment) intersecting D5 (Design Under Constraint) at a moment when the cheaper path was to say nothing.

F6: Handling a Customer Data Exposure During an Outage

A production incident caused an LLM to surface one customer's data in another customer's session. The technically fastest path was to hot-patch the prompt and continue. Christiam insisted on full disclosure, coordinated the incident report with the affected party, and drove the architecture change that prevented recurrence: tenant isolation via separate embedding indexes and retrieval-scoped authorization, even though it delayed the sprint. This is D6 (Professional Judgment) in a high-stakes incident context.

Each scenario maps to the format families defined in AII-CS-003 §3, the Institute's published competency standard.

Teaching Method

From Production to Classroom

Students in Christiam's courses do not work from clean-slate prompts. They inherit broken retrieval systems, underspecified feature requests, and systems with no measurement, because those are the conditions he has operated under professionally. Every exercise he designs maps to one or more domains in the Institute's Competency Taxonomy.

D1: Systems Construction

Students build LLM-integrated applications, RAG pipelines, and agentic architectures from ambiguous requirements: exactly as Christiam has done for clients. The work is judged on correctness, maintainability, and engineering judgment, not completion.

D2: Production Operation

Students deploy to AWS, set up CI/CD pipelines, configure observability with LangSmith, implement rate limiting and semantic caching, and manage cost controls. Christiam brings real cost data from his production systems into the classroom.

D3: Evaluation & Measurement

Students design evaluation harnesses, define quality metrics, and learn to measure what matters rather than what is easy to measure. Christiam teaches that an AI system without measurement is an AI system you cannot improve.

D4: Diagnosis & Improvement

Christiam presents students with systems that are degrading and partially observable. They learn root-cause analysis across the stack - from data drift to prompt degradation to index corruption to infrastructure failure - and must isolate, fix, and document the cause.

D5: Design Under Constraint

Architecture decisions with real trade-offs: latency versus accuracy, cost versus capability, speed versus safety. Students must decide, document, and defend their choices before faculty: the same review process Christiam has faced in production architecture reviews.

D6: Professional Judgment

Scenarios involving data exposure, honest reporting, deployment risk, and escalation. Christiam teaches that technical decisions are also professional decisions, and that the right call is not always the fastest one. This domain carries standalone weight in every assessment.

Areas of Instruction

Courses Christiam Instructs

Christiam teaches across Terms 2 and 3 of the AI Engineering Bootcamp, covering the modules where production AI engineering depth matters most. Each course is part of the published curriculum and is assessed through graded assignments, practical coding assessments, and scenario-based evaluations.

AIE 202

LLM APIs & Prompting

OpenAI SDK Integration
Prompt Engineering Patterns
Parameter Tuning
Streaming Responses
Structured Outputs

AIE 204

RAG Pipelines

Chunking Strategies
Document Parsing
Vector Database Setup (Qdrant)
HyDE Re-ranking
Multi-Modal RAG

AIE 301

LangChain & LangGraph

Chains, Prompts & Memory
Tool Use & Function Calling
State Graphs with LangGraph
Human-in-the-Loop
Multi-Agent Orchestration

AIE 302

Agentic Systems

ReAct Agent Pattern
Tool-Using Autonomous Agents
Planning & Decomposition
Reflection & Self-Correction
Handling Infinite Loops

AIE 303

Production & Deployment

AWS Deployment (ECS / Lambda)
API Design for LLM Apps
Rate Limiting & Throttling
Semantic Caching
CI/CD for AI Pipelines

AIE 304

Observability & Evaluation

LLM Tracing (LangSmith)
Automated Evaluation
Token & Cost Monitoring
A/B Testing LLM Configs
AI Evaluation Methods

Why This Matters

What Students Gain From His Courses

Christiam Ipanaque does not teach from a textbook. Every technique he demonstrates, every architecture pattern he recommends, and every failure mode he warns about comes from a system he personally built, operated, or repaired in a production environment serving real users. His students leave with more than knowledge: they leave with judgment.

Production mindset: how to build AI systems that survive contact with real data, real users, and real failure modes. Christiam has shipped systems handling 12,000+ queries and 15,000+ documents per month. He teaches the engineering discipline that makes those numbers possible.
Diagnostic ability: how to locate problems when observability is broken, a stakeholder is pressing for answers, and the cause could be anywhere from the prompt to the index to the infrastructure. He has done this in production under real pressure.
Trade-off fluency: how to make decisions between capability, cost, latency, and safety under deadline. Christiam has redesigned architectures mid-deployment to meet cost targets without sacrificing quality. He teaches the same framework to his students.
Professional judgment: how to recognize when the technically fastest path is the professionally wrong one. Christiam has caught data-handling flaws in architecture reviews and insisted on disclosure during incidents. He treats judgment as a teachable, assessable competency.
End-to-end ownership: how to own a system from ambiguous requirement to deployed production service with monitoring, evaluation, and cost controls. Christiam founded and runs Northbridge Engineering. He teaches students to think like owners, not ticket-takers.

Standard & Assessment

Built on a Published Competency Standard

Christiam's courses are not graded on attendance or recall. Every student is assessed through the Institute's published scenario-based evaluation method, which is documented openly in the AI Engineer Competency Standard. Students receive a competency profile with a level per domain, rather than a single grade. The results are recorded on an official transcript that employers can verify.

What the Standard Measures

D1: Systems Construction: building correct, maintainable AI applications.
D2: Production Operation: deploying and operating AI systems at scale.
D3: Evaluation & Measurement: making system quality measurable.
D4: Diagnosis & Improvement: finding and fixing degradation in probabilistic systems.
D5: Design Under Constraint: architecting against real-world trade-offs.
D6: Professional Judgment: decisions where technical feasibility and responsibility diverge.

How Students Are Assessed

Scenario-based evaluations that place students inside realistic engineering situations.
Scored on decisions and reasoning, not recall or completion.
Graded by calibrated evaluators against published rubrics.
Every score is auditable from criterion to observation.
Results appear on the official transcript with domain-level granularity.

Christiam Ipanaque: AI Engineer & Instructor