Machine Learning Architect, Personalization Resume Mila Aguirre Machine Learning Architect, Personalization (206) 555-0148 mila.aguirre@example.com linkedin.com/example/milaaguirre Seattle, WA 98109 STRENGTHS Architecture Leadership Unified six recommendation surfaces through reusable interfaces; teams gained clearer ownership and faster technical decisions. Relevance Judgment Balanced recall, precision, NDCG, diversity, conversion, and AOV; product partners gained confidence in releases. Cross-Team Influence Architecture reviews aligned more than 40 contributors without formal authority; difficult tradeoffs became shared decisions. Production Focus Investigated serving issues alongside model teams; practical fixes kept latency, freshness, and reliability visible. Mentorship Research reviews and design critiques helped 12 engineers and scientists make stronger choices under pressure. SKILLS Recommender Systems Retrieval Learning to Rank Reranking Personalization Reinforcement Learning Cold-Start Modeling Python Java SQL PyTorch TensorFlow OpenSearch Vector Databases Feature Stores LANGUAGES English Native 40 Spanish Proficient 30 MY CAREER 15.8 Years Principal Machine Learning Architect at Northstar Marketplaces (4.7 Years) Staff Machine Learning Engineer at BrightLoop Media (3.5 Years) Senior Data Scientist, Recommendations at HarborNine Commerce (3.3 Years) Machine Learning Engineer at VectorPath Systems (2.4 Years) Software Engineer at Cedar Ridge Software (1.9 Years) SUMMARY Principal-level Machine Learning Architect with 16+ years across software engineering, machine learning, and data science , including 11+ years building recommender systems for commerce, media, and marketplace products. Brings hands-on leadership across retrieval, ranking, reranking, real-time serving , feature stores, vector databases, streaming pipelines, and low-latency APIs. Sets technical direction through architecture RFCs, technical reviews, platform migrations, research prototypes, code reviews, and production investigations. Connects engineering, data science, product, merchandising, and business priorities through measurable decisions involving recall, precision, NDCG, diversity, conversion, and AOV. Ready to help LoomCart make secondhand shopping easier through scalable, relevant, adaptable personalization. EXPERIENCE Principal Machine Learning Architect Northstar Marketplaces January 2022 - Present Seattle, WA Owns architecture for personalized marketplace experiences across six surfaces, guiding engineers, data scientists, product leaders, and merchandising partners. Shapes reusable recommendation platforms, investment decisions, service standards, and production practices for an 18-million-item catalog. Established reusable retrieval interfaces across six marketplace surfaces, reducing fragmented recommendation architecture . Authored architecture RFCs for feature ownership, deployment standards, and service-level objectives across 40+ partners. Built OpenSearch candidate retrieval with vector search and streaming features for 18 million active items. Created evaluation gates linking recall, precision , NDCG, diversity, conversion , and AOV for release decisions. Migrated batch features toward continuous refresh, improving cold-start coverage for newly listed products and sellers. Mentored engineers and scientists through design critiques, research reviews, incident analysis, and production debugging. Staff Machine Learning Engineer BrightLoop Media June 2018 - December 2021 New York, NY Led technical direction for personalized discovery across web, mobile, and partner applications. Coordinated retrieval, ranking, experimentation, feature platforms, and stakeholder decisions across engineering, data science, product, and commercial groups. Coordinated retrieval and ranking services supporting personalized discovery across web, mobile, and partner applications. Built two-stage retrieval and reranking with learning-to-rank models, nearest-neighbor search, and online features. Connected NDCG and recall changes with completion rate, conversion, and AOV through online experimentation. Migrated teams toward shared feature definitions and streaming pipelines , making freshness visible across products. Converted catalog quality, merchandising , and revenue goals into model requirements and service interfaces. Mentored 12 engineers and scientists through architecture reviews, incident retrospectives, and implementation planning. Senior Data Scientist, Recommendations HarborNine Commerce January 2015 - May 2018 Austin, TX Developed recommendation models and evaluation systems for a growing commerce catalog. Partnered with product and merchandising teams on experiments, relevance tradeoffs, inventory constraints, and measurable commercial outcomes. Developed personalized retrieval and ranking with implicit feedback, item attributes, session context, and seller signals. Built reusable datasets monitoring precision, recall, NDCG , coverage, and diversity across model releases. Implemented inventory-aware reranking with category constraints, price bands, and merchandising priorities. Designed experiments linking recommendation quality with conversion, basket composition, and average order value. Prototyped Kafka -based feature generation and service caching for recent views, searches, and purchases. Presented model tradeoffs to technical and business audiences, shaping sequenced delivery plans. Machine Learning Engineer VectorPath Systems July 2012 - December 2014 Chicago, IL Implemented recommendation pipelines and retrieval services for retail and media clients. Supported model evaluation, OpenSearch integration, deployment practices, data quality, and client-facing technical delivery. Implemented batch recommendation pipelines with collaborative filtering, content features, and scheduled model scoring. Built retrieval evaluation tools comparing candidate-generation strategies against precision, recall, and ranking benchmarks. Integrated OpenSearch indexes with Python services for fast item lookup and catalog-aware experiments. Added validation and lineage checks across clickstream, catalog, and transaction feeds before model jobs. Containerized inference components with documented deployment, rollback, and troubleshooting procedures. Delivered client demonstrations clarifying cold-start behavior, system limits, and expected business outcomes. Software Engineer Cedar Ridge Software July 2010 - June 2012 Raleigh, NC Built Java and Python services for search, catalog metadata, customer activity, and distributed commerce workflows. Contributed testing, ingestion, relevance, defect investigation, and an early personalized browse prototype. Developed Java and Python services for search, catalog metadata, and customer activity processing. Created batch ingestion jobs with schema validation and retry handling for dependable analytics feeds. Improved search relevance through OpenSearch analyzers, filters, and query templates. Built API tests and performance checks, documenting latency findings for senior reviewers. Investigated production defects with QA, product, and operations partners, translating workflows into maintainable fixes. Presented a personalized browse prototype that launched subsequent recommendation research and experimentation. PROJECTS Real-Time Resale Personalization Platform 2025 Created an adaptive recommendation concept for resale catalogs, combining streaming behavior, vector retrieval, reranking, cold-start signals, and low-latency serving. Focused decisions on relevance, diversity, conversion, and AOV. Open-Source Ranking Toolkit 2019 Maintained reusable ranking components and evaluation utilities for retrieval research. Contributions supported clearer comparisons across recall, precision, NDCG, and reranking approaches. LEADERSHIP & AWARDS ACM RecSys Industry Impact Recognition, 2023 Internal Distinguished Technical Leadership Award, 2021 Open-source Contributor Recognition for Ranking Toolkit, 2019 EDUCATION Master of Science in Computer Science North Carolina State University GPA: 4.0 2010 Raleigh, NC Coursework: Machine Learning, Distributed Systems, Information Retrieval, Data Mining, Software Engineering CERTIFICATIONS AWS Certified Machine Learning - Specialty 2024 Google Professional Machine Learning Engineer 2025 TECHNICAL SKILLS Recommendation Models: Collaborative Filtering, Learning to Rank, Reranking Search Platforms: OpenSearch, Elasticsearch, Query Templates Machine Learning Frameworks: PyTorch, TensorFlow, Scikit-learn Programming Languages: Python, Java, SQL Feature Infrastructure: Feature Stores, Online Features, Feature Engineering Vector Retrieval: Vector Databases, Approximate Nearest Neighbor, Embeddings Streaming Systems: Kafka, Streaming Pipelines, Event Processing Distributed Systems: Distributed Services, High Throughput, Scalability Serving Systems: Real-Time Serving, Low-Latency APIs, Model Deployment Evaluation Methods: Recall, Precision, NDCG Business Metrics: Conversion, Average Order Value, Diversity Cloud Platforms: AWS, Containers, Cloud Services Experimentation: A/B Testing, Online Experiments, Offline Evaluation Architecture Practice: Architecture RFCs, Technical Reviews, Platform Migrations Data Operations: Clickstream Data, Catalog Data, Data Lineage SKILLS Recommender Systems Retrieval Learning to Rank Reranking Personalization Reinforcement Learning Cold-Start Modeling Python Java SQL PyTorch TensorFlow OpenSearch Vector Databases Feature Stores PROFESSIONAL AFFILIATIONS ACM RecSys Conference Volunteer Reviewer, 2022 - Present Women in Machine Learning Mentor, 2018 - Present Open-source Ranking Systems Maintainer, 2019 - Present LANGUAGES English (Native) Spanish (Proficient) ADDITIONAL INFORMATION Work Status : Authorized to work in United States. No sponsorship required. ADDITIONAL INFORMATION Work Status : Authorized to work in United States. No sponsorship required. REFERENCES AVAILABLE ON REQUEST
  • File
  • Preview
  • Show Styles
  • PDF Download
  • US English
File
Styles Download Buy Log In
Bootstrap
>JP 1 1
PDF Footer and Margins Setup
The margins are applied to the document, and the footer shows up in the preview and the PDF

Footer Font Settings

Are you applying for a specific job
so we can tailor your resume to it?Are you applying for a specific job
so we can tailor your cover letter to it?
Type your details as text to generate a personalized resumeType your details as text to generate a personalized cover letter
Website link found! Ready to create resume from:
Skip this Step
Drag the into the box to verify you're human.
Drop here

My Resumes

Displaying {number_of_files} files (out of {total_of_all_files} total)
NameFile TypeLast Modified:Description:

Rename File

Filename already exists!

Delete Files

Warning: This action is irreversible. Once you click submit, the files will be permanently deleted.

Heading text
Upload From Computer Upload
Select From Existing
LinkedIn
Drag & Drop files anywhere in this tab, or

Uploaded Files

DeleteNameLast Updated Use This File
Select this to use the file for generating your resume and cover letter. Only one source—uploaded file, selected file, or LinkedIn text—can be used.

My Files

Resumes or cover letters you create will appear here. No files available yet.
DeleteNameLast Updated Use This File
Select this to use the file for generating your resume and cover letter. Only one source—uploaded file, selected file, or LinkedIn text—can be used.

Add LinkedIn Profile PDF

You can download your LinkedIn profile as a PDF by opening linkedin.com in a web browser, clicking your profile photo, selecting the “More” or “Resources” button, and choosing “Save to PDF.” You can then attach the PDF using the Upload From Computer tab.

Paste Your LinkedIn Profile Text

If you have copied the LinkedIn profile text, paste it below. See how to copy your profile from LinkedIn

   
Select this to use the LinkedIn profile text for generating your resume and cover letter. Only one source—uploaded file, selected file, or LinkedIn text—can be used.
-->
Select this to use the LinkedIn profile text for generating your resume and cover letter. Only one source—uploaded file, selected file, or LinkedIn text—can be used.

Try these controls

<
>