AI Backend Engineer | Generative AI & Multi-Agent Systems

www.linkedin.com/in/michaelcan2/

Manhattan, New York | michaelcan293@gmail.com | (773) 951-8669

I am a Backend Engineer specializing in AI agent orchestration, generative AI, and data engineering. At Annalect (Omnicom Media Group), I design and deploy LangChain based multi-agent systems, FastAPI microservices, and MCP server infrastructure serving 6,000+ users. My work has delivered measurable impact, including reducing Google Ads campaign time-to-market by 50%, increasing click-through rates by 25%, and improving campaign brief completion through guided AI workflows. I hold an M.S. in Artificial Intelligence from Northwestern University and certifications in Azure AI and Data Fundamentals.

Skills

• Languages & Frameworks: Python, SQL, FastAPI, REST APIs, LangChain
• AI Platforms & Tools: OpenAI, LLMs, RAG, MCP Servers, NLP
• DevOps & Tools: Docker, Streamlit
• Cloud: AWS (Lambda, EventBridge, API Gateway, S3), Databricks, Azure (Functions, Blob, Cognitive Search, DevOps)

Professional Experience

Applied AI Backend Engineer (AI / Prompt Engineer) Annalect – Omnicom Media Group (July 2024 - Present) New York, NY

AI Agent Orchestration & Marketing Optimization – LLM Routing, FastAPI Microservices, and MCP Server Deployments

• Developed a LangChain-based agentic orchestrator that dispatches user prompts to modular sub-agents (Web Search, File Retrieval), dynamically invoking specialized MCP tools via LLM-driven routing to serve 6,000+ users.
• Built a multi-agent Audience Builder workflow converting campaign intent into attribute based audience profiles by filtering a global population across thousands of Omni data signals to produce targetable, insight-rich segments.
• Integrated OpenTelemetry based tracing with an LLM driven evaluation framework, using LLM as a judge techniques to score agent output quality, and RAG relevance, against golden datasets.
• Built an Influencer Insights MCP leveraging Databricks Genie and Unity Catalog, enabling media planners and strategists to query influencer and audience data in natural language and translating intent into validated SQL against trusted Databricks datasets.
• Led development of a Next Gen Search AI Agent for Google Ads keyword optimization, adopted by 50+ client accounts, improving time-to-market by 50% and click-through rate by 25%.

AI Engineer / Software Engineer @ PricewaterhouseCoopers (PwC) (December 2022 - March 2024) New York, NY

Leading Legal Insurance Company – Azure Services NLP Machine Learning Unitization and Metadata Extraction pipeline

• Engineered a reusable code accelerator for an NLP pipeline, transforming OCR data into structured output integrating different off-the-shelf components for each processing step through the Azure cloud services platform.
• Built multiple Azure Functions (equivalent to AWS Lambda) to transfer client documents through an Azure Service Bus messaging queue pipeline, facilitating multiple transformations from PDFs to text to JSON document models containing metadata extracted from the specific documents.
• Established the Microsoft Azure Cognitive Search service, fully implementing it with unitized documents and extracted data. This allows desired metadata to be immediately accessible through search/filter queries.
• Led Agile sprints to achieve agreed-upon objectives, coordinating and executing corrective actions to address gaps with the client.
• Developed a text classification algorithm using the spacy Python library to automatically detect entities and key topics within unstructured financial text data and performed further NLP text analysis. Created a Python script for converting OCR documents into text.

AI Factory Gen AI Functional Specification Documents (FSD) to SQL conversion – Data Engineer

• Developed a pipeline using Apache Spark Databricks notebooks in Python that takes Functional Specification Documents (FSD) as input and generates a usable SQL query using OpenAI API.
• Used Azure Cognitive Services (ACS) to store preexisting Functional Specification Documents (FSDs), enabling the retrieval of their corresponding SQL queries. These SQL queries were then utilized as input for the LLM, which generated additional SQL based on the given input.
• Pipeline reduces development time by ~6 hours, automating the extraction and insertion of information from FSD documents into SQL according to sponsors.

Test Case Generation – Gen AI Developer / Data Engineer

• Crafted a comprehensive technical architecture illustrating each facet of the pipeline and its operational workflow.
• Established Retrieval Augmented Generation (RAG) infrastructure integrated with Azure Cognitive Search (ACS) to serve as a robust knowledge base, ensuring the availability of high-quality test cases for reference when generating new test cases for upcoming FSDs.
• Employed chunking and embedding methodologies to segment multiple test cases in Excel, facilitating their integration through a standardized ADA-embedding model prior to populating them into the ACS index.
• Managed the Test Case Generation Pod during Agile sprints with Azure Dev Ops (ADO) by establishing features, user stories, and assigning tasks.

Gen AI Procurement Chatbot MVP – Gen AI Developer / Data Engineer

• Implemented a Procurement MVP chatbot utilizing Langchain library with the GPT-4 LLM Model.
• Deployed Minimum Viable Product (MVP) as a secure Webapp through Docker.
• MVP saves ~8 hours of time eliminating procurement team need to invest in incomplete requests.
• Prompt engineered MVP tool to educate users on all the necessary information before they can send requests, based on client feedback, to guide users in creating requests.

AI Engineer / Software Engineer @ Sagence Consulting (March 2020 - November 2022) New York, NY

Large Academic Institution – AWS Snowflake Architecture

• Built a reusable code accelerator pipeline for the firm for ingestion of the client’s data from the Client Relationship Management (CRM) Salesforce organization into Snowflake database utilizing AWS services such as API Gateway, EventBridge, Lambda, S3 through CloudFormation.
• Stood up API Gateway to receive data from CRM; launched an EventBridge to create a validator for the data, ensuring it meets the database’s table structure when updates occur; implemented a Lambda function that filters the JSON data received from EventBridge to specific data points the client defined as resourceful and placed them into S3.
• Instructed the Snowflake database to use Snowpipe for ingesting preprocessed data into the expected table within Snowflake once S3 was populated with new data.

Large Financial Holding Company – Enterprise Data Architecture

• Constructed requests to an outdoor vendor API through Apache Databricks, and scheduled jobs to establish a Data Ingestion pipeline for incoming data in a distributed system.
• Analyzed, sanitized, and developed Fact and Dimension tables that populate the Insurance Company’s new Data Warehouse.
• Automated ETL processes as part of the Company’s ingestion system to update Data Warehouse.

Education

Northwestern University, McCormick School of Engineering March 2020 Master of Science in Artificial Intelligence Evanston, IL

Coding Dojo, Coding Bootcamp July 2017 Fundamentals of Programming in Python Chicago, IL

Columbia College Chicago May 2015 Bachelor of Arts in Television Chicago, IL

Certifications

Microsoft Certified: Azure AI Fundamentals Issued February 2024 Demonstrating proficiency in machine learning and AI principles. Credential ID: 84BE9C305FAB7BBD

Microsoft Certified: Azure Data Fundamentals Issued January 2024 Demonstrating expertise in data principles and Azure data solutions Credential ID: 54DA70C0BFB87218