Yun Huang

Work
- I study how intelligence behaves under pressure, how models drift when emotion, authority, or incentives test their reasoning. I’m interested in what behavioral shifts reveal about alignment, safety, and the evolution of machine reasoning
- Before research, I built monetization and ML infrastructure at LinkedIn and Jet.com ($3.3B acq.), where I learned how incentives shape both markets and models. I recently lectured in Penn’s machine-learning application course, discussing how firms weigh rule-based vs ML features when mitigating online price wars
- Please see full work milestones below
‍
Personal
- I live in San Francisco, I'm a U.S. Citizen. Native fluency in English, Cantonese and Mandarin
- Writers I admire include Noah Smith, Nick Bostrom, and Immanuel Kant. I enjoy science fiction films which provoke thoughts about the complexities of human life, such as The Three-Body Problem and The Prestige. I admire Marie Kondo's way of life, some of my favorite brands include Gentle Monster and Totome
- Beyond primary work, I run, take photos, make coffee, and think about how humans (and machines) reason about one another
‍‍
Interests & Personal Philosophy
- Data: I believe the demand for high-quality training data is underestimated, and we will see the rise of data brokerage markets for AI
- Compute: Compute will drive not only decisions on what should be built, but what actually ships
- AI Research: I follow work on alignment and interpretability research
- Application: I'm tracking how core research labs vs. independent entities will faire in scaling real-world commercial adoption
- Life: I believe that your environment—especially your social and informational networks—shapes who you become more than your conscious choices do. What feels like “intentionality” is often just the emergent effect of subtle, repeated training signals from the people and structures around you

Work Milestones Full List
- 10/2025: Hosted a model behavior salon across labs to brainstorm drift metrics and iteration criteria for next run (i.e. prompt-set expansion across domains, prompt variations from 6 to 20 paraphrases)
- 08/2025: Created one of the first behavioral drift benchmarks, evaluating GPT-5 vs. 4o (72 hours post GPT-5 launch), subsequently comparing GPT-5 vs. Claude 4.1 (2 weeks post launch)
- 05/2025: Guest Lecturer, Machine Learning Application, delivered a lecture to 500+ students in Penn's “Analytics & Machine Learning Application” curriculum, focused on trade-offs between rule-based vs. ML features in mitigating online price wars
- 03/2024: AI Adoption at LinkedIn, initiated creation of one of the organization's first AI prompt libraries, adopted by 300+, generated 20 initial use cases (~75% for automation, 25% for creativity). Saved each ~15 hrs/quarter
- 06/2024: Early identifier of Ukraine War advancement, led 600+ in creating spend pause signals during the Ukraine War. Detected pauses across ~30% of accounts, improved forecasting accuracy by ~15%, prevented ~$xM in quarterly revenue loss
- 05/2024: Lead Scoring Product (ML): Collaborated on launching ML-enabled lead-scoring product (Including sourcing, integration of Pathmatics), increasing conversion amongst adopted teams by 30%, generating ~$xM incremental revenue
- 01/2023: Joined LinkedIn to drive monetization and ML-enabled initiatives for fastest-growing vertical (> $xB Advertising unit)
- 08/2021: Promoted to Pricing VP Chief of Staff after 6 months in seat. Drove $40B GMV roadmap across 30+ in Product, Data Science and Engineering
- 10/2020: Purchased a home in the Greater New York area
- 02/2020: Collaborated with ML Data Scientists in testing rule-based vs. ML features in mitigating online price wars; enforced rule-based approach, achieving 2× profitability for 1/5 of anticipated resourcing. Developed anomaly detection dashboards, identifying ~70% of falsified use cases, mitigating ~$xM losses
- 02/2020: Invested in real estate in the Greater New York area
- 09/2018: Joined Jet.com ($3.3B acq.), managing $120M P&L. Launched one of the company's first inventory analytics products, decreasing access time for hundreds from days to minutes; increasing supply rate from 80% to 95% in weeks. <1% to self-select entry point with VP-team advocacy, <5% to receive accelerated promotion (9 months) from Associate to Senior Associate, and <5% from function to transition to Dynamic Pricing product as the youngest team member (ex-McKinsey, BCG exits). Women's ARG Co-lead, led speaker series (i.e. Marc Lore, Andy Dunn, Jenny Fleiss) serving 300+ members
- 2017: Performed research under Professor Maurice Schweitzer at Wharton Behavioral Labs, focused on how emotions impact decision behaviors
- 2017: Co-developed one of The New York Times’ first short-audio features for “The Daily,” increasing engagement, generating ~$xM incremental revenue
- 2014–2018: University of Pennsylvania, B.A. in Economics & Psychology (Focus: ML Application). Awarded full $240K merit-based scholarship; Dean’s List
‍