Lindsey Ma 马兰馨

I'm a recent M.S. graduate in Statistics from Columbia University, with a focus on computer vision and machine learning. I'm also an avid cyclist & hiker.

Previously, I received my B.A. in Economics from Renmin University of China.

Experience

I build and deploy machine learning systems for computer vision.

Mobile plant recognition and species identification system

Plant Recognition System

SWE Intern, Arborphy AI
May 2026 – Present, New York, ongoing

Developing a dual-encoder architecture combining EfficientNet and BioCLIP through cross-attention feature fusion to integrate organism- and species-level representations while addressing negative transfer.

Investigating representation failures using frozen-embedding clustering, linear probes, and rotation-based local projections, while exploring active learning-guided fine-tuning to improve classification performance.

Distilling a lightweight student model for offline iOS inference and implementing cloud fallback mechanisms for ambiguous predictions.

Feature Fusion · Representation Learning

Real-time player pose and shuttlecock detection for a badminton training system

Real-Time Vision for Robotics

Software Developer, InnoX Academy Robotics Lab
May 2025 – December 2025, Shenzhen

I built a 30 FPS real-time vision pipeline on Raspberry Pi for a badminton training robot, integrating pose estimation, shuttlecock tracking, and court positioning to drive adaptive serving and technique feedback.

I deployed a local LLM voice interface with Ollama and onboard audio for hands-free robot control.

Computer vision · Edge deployment

Research

In Liam Paninski’s lab, I worked on recovering neural signals from noisy two-photon calcium imaging.

Two-photon calcium imaging data used for self-supervised denoising

Self-Supervised Learning for Neural Imaging

Research Assistant, Center for Theoretical Neuroscience, Columbia Zuckerman Institute
January 2026 – May 2026, New York

Designed and trained a self-supervised blind-spot neural network in PyTorch for denoising two-photon calcium-imaging movies under temporally correlated noise.

Conducted systematic ablations of masking strategies, receptive fields, and preprocessing across five real and simulated imaging datasets, evaluating latent-signal recovery and neuronal morphology preservation.

Rebuilt the lab’s PMD preprocessing pipeline, reducing residual video energy by 20% and downstream decoding variance by 7% versus baseline.

Computational neuroscience · Self-supervised learning · Two-photon calcium imaging

Selected Projects

Three-dimensional UMAP embedding of the FlyWire connectome

Community Structure in the FlyWire Connectome

Computational Statistics Project, May 2025

Applied Bayesian stochastic block models to the FlyWire whole-brain connectome to infer cell-type organization; evaluated community structure with posterior predictive checks and visualized embeddings.

Bayesian modeling · Connectomics

Common spatial pattern maps used for EEG motor-imagery decoding

Motor-Imagery Decoding from EEG

Brain Computer Interfaces Lab Project, January 2025

Extracted spatial features from real-time EEG recordings using Common Spatial Pattern (CSP) and trained a linear discriminant analysis (LDA) classifier to decode left- versus right-hand motor imagery, achieving 80% mean accuracy.

Brain computer interfaces · EEG signal processing

CAD design, simulation, and physical prototype of a legged robot

Multi-Joint Robot Design and Simulation

Robotics Studio Project, January 2025 – May 2025

Designed and built a multi-joint robot from CAD modeling and 3D printing to motor integration and physical assembly; developed a physics-based simulation and validated individual joint actuation.

CAD · 3D printing · Robotics simulation

Hackathons & Technical Activities

Blockchain Hackathons, Team Leader

September 2025 – October 2025

Won second place at the XRPL Hackathon ($5,000) and third place at the Stellar Hackathon ($3,000). Led the team to build StrataFi and venMorph in separate 24-hour hackathons, implementing smart-contract integrations, backend transaction-processing pipelines, and multi-wallet testing suites.

Columbia AI Alignment Club, Technical Fellow

February 2025 – May 2025

Reproduced Anthropic’s hallucination-circuit analysis on Qwen-8B using transcoder-based attribution graphs; applied feature-guided activation steering to induce reproducible refusal-to-hallucination shifts.

Education

Columbia University

M.S. in Statistics
September 2024 – May 2026
GPA: 3.71/4.00

Neural Networks & Deep Learning, Advanced Machine Learning, Computational Statistics, Time-Series Modeling, Stochastic Processes, Statistical Analysis of Neural Data, Robot Learning, Robotics Studio, Brain Computer Interfaces Lab.

Renmin University of China

B.A. in Economics
September 2019 – June 2023
GPA: 3.82/4.00

Technical Skills

Programming & ML

Python, R, SQL, PyTorch, Hugging Face, scikit-learn

Systems & Tools

FastAPI, Docker, vLLM, Ollama, Neo4j

Areas of Interest

machine learning systems, computer vision, computational neuroscience