Akshita Arora

Machine Learning Engineer @ US HealthCenter, Inc.

About

I build ML systems that go into production and stay there.Currently an ML Engineer at a healthcare AI company while finishing a dual master's (Machine Learning / ECE + Mechanical Engineering) at the University of Michigan (4.0 GPA). Before that, 3.5 years as a Scientist at ISRO, where my models ran on human spaceflight hardware.What I actually do:- Ship end-to-end pipelines: data ingestion → feature engineering → inference APIs serving predictions at <200ms in clinical decision systems- Build deep learning models for high-stakes domains: disease risk stratification (0.81 AUC), physical property inference from multimodal sensors (−15% error vs. baselines), anomaly detection on propulsion systems (+30% accuracy)- Work at the intersection of ML research and engineering: I've published at 3 international conferences and written the production codeI'm drawn to AI/ML startups moving fast on hard problems in scientific, healthcare, or applied AI domains, where the gap between research and deployed product is where the real work happens.Open to: ML Engineer, Applied Scientist, and AI/MLOps roles.

Country

United States

City

San Francisco Bay Area

Industry

Computer Software

Skill

Artificial Intelligence (AI), ML, Pandas (Software), SciPy, Algorithm Development, Data Structures, Machine Learning Algorithms, Mathematics, Design Failure Mode and Effect Analysis (DFMEA), New Concepts, Problem Solving, Teamwork, Communication, Simulation Software, Hardware Testing, Hardware Design, Verification and Validation (V&V), Modeling, Computer Science, Prototyping

Experience

Machine Learning Engineer

US HealthCenter, Inc.

2025-5 - Present · 1 yr 5 mos

United States

Built the ML infrastructure powering clinical risk prediction across a patient population of 100k+. • Reduced model iteration time by 40% by designing modular Python pipelines covering ingestion, feature engineering, training, and inference: supporting 8+ reproducible experiments with full audit trails for regulatory review. • Shipped production inference APIs integrated with downstream clinical decision systems, serving real-time predictions at <200ms latency. • Delivered risk stratification models achieving 0.81 AUC on holdout sets, directly informing care prioritization for high-risk patients. • Translated legacy SQL-based algorithms into scalable Python for deployment: eliminating a bottleneck between data science and engineering teams.

ML Research Assistant

University of Michigan College of Engineering

2025-3 - 2025-6 · 4 mos

United States

Researched deep learning methods for inferring physical system properties from multimodal sensor data: optical, spectral, and numerical inputs combined. • Achieved 15% lower prediction error vs. published baselines by designing PyTorch/JAX models that fused heterogeneous sensor modalities. • Ran systematic experiments across 20+ architecture and hyperparameter configurations, with interpretability and robustness analysis at each stage. • Built validation pipelines across 5 operating regimes, surfacing failure modes not captured by existing monitoring approaches.

Scientist/Engineer SC

ISRO - Indian Space Research Organization

2021-1 - 2024-7 · 3 yrs 7 mos

Thiruvananthapuram

One of a small team of engineers responsible for the ML and thermal systems on the Crew Escape System (CES): India's first human-rated launch vehicle (Gaganyaan). • Improved propulsion anomaly detection by 30% by applying ML and statistical modeling to data from 200+ spaceflight system tests. • Built neural network models predicting solid rocket motor throat erosion, achieving 88% reliability accuracy: results used to inform maintenance and design decisions. • Contributed to Probabilistic Risk Assessment (PRA) studies for human-rated launch systems; analysis influenced 3 accepted design change recommendations. • Missions: TV-D1 (Test Vehicle Abort Mission-1), Gaganyaan crewed spaceflight program.