I'm Bryan, I was born in Yangon, Myanmar and raised in San Francisco, CA.
My background is in mechanical engineering and manufacturing. I graduated from Cornell University with B.S. in Mechanical Engineering. At Apple, I worked on product design, validation, and failure analysis for the iPhone 17, iPhone 17 Air, and iPhone 15 Pro Max. After leaving Apple, I designed a robotic system for performing maintenance on autonomous vehicles. Shortly after that, I founded Reframe, an AI procurement tool for mechanical hardware, as part of Y Combinator’s W26 batch.
With my team at Apple Park.
Outside Y Combinator.
Hiking at Pali Notches in Hawaii.
On a dive boat at the Similan Islands in Thailand.
A moray eel on a dive.
1 / 5
Currently, I work on machine learning and empirical analysis surrounding questions about supply chains, firms, and markets. My current research measures supply-chain risk in corporate disclosures and examines its relationship with stock-market reactions. My projects also examine how modeling and evaluation choices affect prediction, from chronological validation to numerical optimization.
I am seeking research associate and fellowship opportunities in applied machine learning and operations research as I prepare for doctoral study.
Outside of work, I like to travel, both within the United States and internationally. I enjoy hiking in national parks, exploring local towns, and going scuba diving. At home, I play basketball and badminton, and I'm learning to make pottery. :)
Selected work
Supply-chain language and market reactions: A 2010–2019 study combining dictionary-based NLP, event studies, and an investigation of how tied scores affect portfolio comparisons. Research overview
Predicting Hacker News engagement: Chronological evaluation of prediction models on a corpus of 4.7 million posts, with an emphasis on data quality and information available at submission. Selected projects
Reconstructing Adam experiments: NumPy implementations of five optimizers and manual backpropagation, including an investigation of results that differ from the original paper. Code and experiments
Research
Research
Supply-chain language and stock-market reactions
Research direction
I am interested in how firms make operational decisions under uncertainty, how disruptions move through production networks, and what can be learned about these processes from text and market data. My current work focuses on measurement and empirical evaluation. Longer term, I want to develop computational models that help evaluate how operational and policy decisions affect firms and markets.
Supply-chain risk language and stock-market reactions
Evidence from US hardware firms · Independent research · 2026–present
Does the way managers discuss supply-chain problems convey information associated with stock-market reactions? Motivated by my experience in hardware product development, I study this question using 11,950 earnings calls from 378 US hardware firms covering fiscal years 2010–2019.
Building on Theile et al. (2026), I construct dictionary-based measures of supply-chain risk and resolution language, using a vocabulary derived from 50,000 annual filings. The analysis combines Carhart four-factor abnormal returns, portfolios that allocate tied scores proportionally, and regressions with firm and calendar-quarter fixed effects.
Calls in the highest-risk portfolio have two-day abnormal returns 1.69 percentage points below those in the lowest-risk portfolio. With fixed effects, a one-standard-deviation increase in risk language is associated with a 0.82 percentage point lower return. On the 5,476-call subsample with earnings-surprise data, adding a provider-reported surprise control reduces the estimate from −0.85 to −0.71 percentage points. Evidence for resolution language is weaker.
Mean CAR(0,1) by fractional SCRisk portfolio, fiscal years 2010–2019. Bars show 95% confidence intervals clustered by firm. Sample: 11,950 calls from 378 US hardware firms. Portfolios share observations when scores are tied.
These findings describe associations: other earnings news, sample selection, and event timing limit causal interpretation. My ongoing work examines whether the relationship is stronger for firms that depend on critical suppliers they cannot readily replace.
Independent project · Baseline evaluation complete; embedding experiments in progress
How much of a post’s eventual engagement can be predicted using information available when it is submitted?
I designed and orchestrated the project. The Python pipeline processes 4,739,207 usable Hacker News stories and identified periods where archived scores reflected submission-time snapshots rather than mature outcomes. After excluding those periods, the baseline evaluation uses 3,568,252 training posts from 2006–2022 and 599,937 test posts from 2024–2025. Historical author and domain features use earlier observations with a score-settling lag.
I compared simple historical baselines, Ridge regression, and XGBoost. The combined-feature Ridge model achieved a test Spearman correlation of 0.297, compared with 0.050 for the trailing-mean baseline. RMSE on log-transformed scores improved by about 3.6%, showing that the features helped ranking more than prediction of score magnitude. XGBoost remained behind Ridge on ranking after early stopping.
The project also includes PyTorch implementations of CBOW and skip-gram objectives with negative sampling, with targeted training experiments on text8. The full embedding variants and their downstream Hacker News comparison remain unfinished.
Independent implementation study · MNIST experiments complete
I implemented Adam, AdaGrad, RMSProp, AdaDelta, and SGD with Nesterov momentum in NumPy, together with manual forward and backward passes for logistic regression and a multilayer perceptron.
The experiments revisit optimizer comparisons from Kingma and Ba’s Adam paper. In my 200-epoch MNIST experiment with dropout, AdaGrad reached a lower final training loss than Adam, differing from the paper’s reported ordering. I documented that discrepancy and possible explanations involving the tuning horizon and gradient sparsity; these remain hypotheses rather than established causes.
The project demonstrates numerical implementation, experiment design, and diagnosis of reproduction gaps. The CIFAR-10 extension is unfinished.
Methods demonstrated in projects: Text processing, dictionary-based NLP, chronological model evaluation, feature engineering, linear and tree-based prediction, numerical optimization, manual backpropagation, and event-study analysis.
Libraries and tools: NumPy, pandas, PyTorch, scikit-learn, XGBoost, Git, and automated testing.
Additional study: Reinforcement learning, Markov decision processes, and policy optimization through independent coursework.
At Reframe, we built procurement automation for hardware teams, connecting bill-of-materials planning, supplier communication, sourcing, and purchasing workflows. The application uses TypeScript, Supabase, TanStack Query for server state, Jotai for UI state, and Trigger.dev for Gmail synchronization and background workflows.
We evaluated negotiation behavior by labeling supplier conversations, documenting effective and ineffective tactics, and constructing preference pairs. We used these pairs to fine-tune gpt-oss with direct preference optimization using MLX.
Production dashboard with assemblies, parts, and status tracking.
Production schedule showing component lead times and project deadlines.
1 / 2
Freshfleet Robotics
Founder · August 2025–November 2025
At Freshfleet Robotics, I designed compliant end effectors for cleaning vehicle seats and floors with a UR10e robot arm, including the cleaning mechanism, electronics, and actuation.
Cleaning hardware and control: I built functional cleaning prototypes with custom end effectors, electrical systems, and circuit design. I developed microcontroller firmware for actuation and hardware control, achieving a three-minute cleaning cycle per seat.
Robot motion planning: I implemented real-time path planning and dynamic collision avoidance in ROS 2 for the UR10e robot arm, and managed an external software contractor to accelerate robotics programming and optimize motion planning.
Prototype funding and deployment planning: I secured $200,000 in pre-seed funding and led deployment negotiations with fleet operator Moove to validate technical specifications against operational constraints.
CAD overview of the cleaning end effector.
End effector mounted on the robot arm.
Test run on a Tesla seat.
1 / 3
Apple
Product Design Engineer · January 2022–May 2025
I worked on mechanical design, structural integration, and validation for iPhone 15 Pro Max, iPhone 17, and iPhone 17 Air. I developed validation protocols, analyzed yield and failure data, and worked with engineering and manufacturing teams across the United States, China, India, and Singapore.
iPhone 15 Pro Max, display integration: I designed and tested the frame integrating the OLED display into the titanium housing. I investigated display delamination and quality defects during engineering development, using CT scans, time-of-flight simulations, Fourier-transform infrared spectroscopy (FTIR), and finite element analysis (FEA) to guide corrective actions.
iPhone 17, antenna module: I designed and tested the antenna module, working on components for Wi-Fi, GPS, 5G+, NFC, and Bluetooth. I addressed bandwidth variation and cellular signal strength through clad-housing design changes and tolerance analyses of grounding clips, electrical cables, and shielding components.
iPhone Air, sealing and validation: I engineered thin adhesive structures to achieve IP68 water and dust resistance within the device’s height constraints. I developed pneumatic pressure and leak-testing protocols to validate seal integrity under thermal and mechanical stress cycling.
iPhone 15 Pro Max display assembly.
iPhone 17 internal antenna components.
iPhone Air internal sealing.
1 / 3
Meta
Product Design Engineer Intern · June 2021–August 2021
At Meta, I designed an experimental privacy slider for Oculus Quest 2, including plastic enclosures, injection-molding considerations, and tolerance stackup analyses. I conducted ergonomics research with more than 20 participants and presented prototype findings to industrial design teams, enabling a 20% reduction in the privacy switch's footprint. I also ran thermal and humidity tests, collecting more than 20,000 telemetry data points for environmental modeling.
Barclays
Summer Analyst · June 2020–August 2020
At Barclays, I evaluated credit line reduction strategies across Barclaycard portfolios using model analytics, regulatory standards, and commercial exclusion policies. I analyzed operational and policy differences across the U.S., Germany, U.K., and Small Business card divisions, and refined targeting parameters for U.S. credit reduction strategies, reducing monthly net credit losses by $13.4 million in July 2020.
BAE Systems
Mechanical Engineering Intern · August 2019–December 2019
At BAE Systems, I designed a hardware interface connecting APKWS electronics with a Honeywell inertial measurement unit, using Creo Sheetmetal for the flex circuit layout. I ran structural and buckling simulations in ANSYS and developed computational fluid dynamics models in STAR-CCM+ to evaluate aerodynamic profiles and support production cost reduction.
Patent
Handheld Electronic Device · US 20240064220 A1
Patent publication covering mechanical architecture and structural integration of the display module into a mechanical housing.
Bryan Zin
