Li Zheng
Data Science | Product | AI
SUMMARY
Data science tech lead with 9+ years of experience shipping data-driven products from 0→1 and scaling them to global audiences. Expertise in experimentation, causal inference, LLM/AI evaluation and integration, search relevance, measurement, and strategic planning.
Experience
Staff Data Scientist, Tech Lead
Meta (Instagram / Messenger) • Sep 2021 - Jan 2026
Led IG Search × MetaAI integration by enabling LLM-based query understanding and personalized ranking; drove the global launch of E2EE on Messenger. IG Search × MetaAI • E2EE
- 0→1: drove and unblocked the integration of AI search into Instagram
- 1→100: rebuilt the search flow (query understanding, retrieval, ranking, personalization & model evaluation) from data foundation to directions
- Designed eval frameworks and benchmark suites for LLM quality, safety, and regression; built human-in-the-loop annotation pipelines and A/B test framework for system-level prompts
- Led a cross-functional analytics team improving IG search quality while enabling MetaAI features such as AI summarization and content recognition
- Drove product go/no-go launch decisions by designing experiments, defining goals/metrics, designing rollout plans, & iterating with causal studies
- Designed customized A/B test frameworks for cluster/network experiments, later became standard practice across Org
- Set team roadmaps by identifying user frictions, sizing/prioritizing opportunities, and estimating timeline; supported DS and engineering hiring
Co-Founder
Flat Strategy • Mar 2025 - Present
Empowering modern homebuyers with expert transaction services, leveraging in-house agentic AI to deliver unmatched transparency, speed, and financial clarity. Check us out!
Senior Associate, Data & Analytics Insights
PwC • Jun 2018 - Sep 2021
Led top banks developing measurable insights, testing and implementing new ideas, mitigating risks, and reducing frictions through advanced quantitative analytics.
- Led analytics engagements for major banks to test/implement new ideas, mitigate risk, and reduce operational friction through quantitative modeling
- Developed and tuned financial risk models and classifiers; ran hypothesis testing and monitoring, saving 100+ hours/week in manual review
- Built an NLP transaction screening system in Python (spaCy, scikit-learn, TensorFlow) and supported deployment across multiple clients
Summer Data Analyst
Duke Health • Mar 2018 - Jun 2018
Optimized fraudulent identification workflow for one of the largest hospital groups in North America, reducing false positive rate from 37% to less than 10%.
- Designed and implemented fraud and anomalies detection workflow
- Built ETL data pipelines, performed DQ analysis, and drafted data dictionaries
Data Analyst
HER Realtors • Aug 2015 - Dec 2016
Developed a real estate price prediction and recommendation system to help investor clients, leading to $5.3M combined deals.
- Delivered model outputs via Tableau dashboards and iterated models/features with stakeholders
- Built and trained machine learning models based on past transaction data
Education
M.S., Data Science (Financial Fraud Analytics)
Duke University • 2017 - 2018
B.S., Accounting & MIS
The Ohio State University • 2012 - 2016
Activity
Private Pilot License (PPL)
FAA • #3877290
Li Zheng
Li Zheng
Data Science | Product | AI
SUMMARY
Data science tech lead with 9+ years of experience shipping data-driven products from 0→1 and scaling them to global audiences. Expertise in experimentation, causal inference, LLM/AI evaluation and integration, search relevance, measurement, and strategic planning.
Experience
Staff Data Scientist, Tech Lead
Meta (Instagram / Messenger) • Sep 2021 - Jan 2026
Led IG Search × MetaAI integration by enabling LLM-based query understanding and personalized ranking; drove the global launch of E2EE on Messenger. IG Search × MetaAI • E2EE
- 0→1: drove and unblocked the integration of AI search into Instagram
- 1→100: rebuilt the search flow (query understanding, retrieval, ranking, personalization & model evaluation) from data foundation to directions
- Designed eval frameworks and benchmark suites for LLM quality, safety, and regression; built human-in-the-loop annotation pipelines and A/B test framework for system-level prompts
- Led a cross-functional analytics team improving IG search quality while enabling MetaAI features such as AI summarization and content recognition
- Drove product go/no-go launch decisions by designing experiments, defining goals/metrics, designing rollout plans, & iterating with causal studies
- Designed customized A/B test frameworks for cluster/network experiments, later became standard practice across Org
- Set team roadmaps by identifying user frictions, sizing/prioritizing opportunities, and estimating timeline; supported DS and engineering hiring
Co-Founder
Flat Strategy • Mar 2025 - Present
Empowering modern homebuyers with expert transaction services, leveraging in-house agentic AI to deliver unmatched transparency, speed, and financial clarity. Check us out!
Senior Associate, Data & Analytics Insights
PwC • Jun 2018 - Sep 2021
Led top banks developing measurable insights, testing and implementing new ideas, mitigating risks, and reducing frictions through advanced quantitative analytics.
- Led analytics engagements for major banks to test/implement new ideas, mitigate risk, and reduce operational friction through quantitative modeling
- Developed and tuned financial risk models and classifiers; ran hypothesis testing and monitoring, saving 100+ hours/week in manual review
- Built an NLP transaction screening system in Python (spaCy, scikit-learn, TensorFlow) and supported deployment across multiple clients
Summer Data Analyst
Duke Health • Mar 2018 - Jun 2018
Optimized fraudulent identification workflow for one of the largest hospital groups in North America, reducing false positive rate from 37% to less than 10%.
- Designed and implemented fraud and anomalies detection workflow
- Built ETL data pipelines, performed DQ analysis, and drafted data dictionaries
Data Analyst
HER Realtors • Aug 2015 - Dec 2016
Developed a real estate price prediction and recommendation system to help investor clients, leading to $5.3M combined deals.
- Delivered model outputs via Tableau dashboards and iterated models/features with stakeholders
- Built and trained machine learning models based on past transaction data
Education
M.S., Data Science (Financial Fraud Analytics)
Duke University • 2017 - 2018
B.S., Accounting & MIS
The Ohio State University • 2012 - 2016
Activity
Private Pilot License (PPL)
FAA • #3877290
Projects
Also available as a Claude Code skill
Data science tool for designing and running statistical experiments
Metrics creation, management, and reporting platform with native AI integration evolved from Meta's internal toolkit: QE, Deltoid, and M360
Modern home buying experience with reduced cost through flat-fee transaction services
AI-powered financial intelligence dashboard that uses LLMs and real-time data to run modular analyses
iOS game featuring AI opponent and local multiplayer
Minimalist productivity timer for focused work sessions