Ashkan Dehghani Zahedani
Fremont,CA
Summary
Senior Machine Learning Engineer and Data Scientist with 5 years of experience at January.ai, Specializing in Time-Series Forecasting, Meta-Learning, and Health AI. Architected the CGM-0 model, a zero-shot learning system that predicts glucose trajectories without hardware data. Published lead author research in Nature Digital Medicine and developed proprietary RNN/LSTM architectures trained on 80,000+ days of multimodal data. Expert in bridging deep learning with clinical validation, delivering scalable API endpoints for enterprise partners like Nestle and Mars.
Machine learning professional with solid history of developing and refining predictive models to solve complex problems. Adept at collaborating with cross-functional teams to achieve impactful results, while remaining adaptable to evolving project needs. Renowned for analytical skills and proactive approach to problem-solving.
Experienced with developing and deploying machine learning algorithms that drive business insights. Utilizes statistical analysis and data mining techniques to enhance model accuracy and performance. Track record of integrating machine learning solutions into production environments, ensuring scalability and reliability. Knowledgeable Data Scientist with robust background in machine learning and algorithm development. Proven track record in designing and implementing advanced models that optimize performance and enhance decision-making processes. Demonstrated expertise in data analysis and feature engineering, consistently driving innovation and efficiency. Professional with strong foundation in machine learning and data science, prepared to drive impactful results. Expertise in developing and deploying machine learning models, optimizing algorithms, and utilizing tools like Python, TensorFlow, and PyTorch. Known for excellent team collaboration and adaptability to evolving project needs. Proven ability to solve complex problems, deliver reliable solutions, and contribute effectively to team objectives.
Overview
5
5
years of professional experience
Work History
Senior Machine Learning Engineer / Data Scientist
January AI
Menlo Park, CA
12.2020 - 01.2026
- Core Algorithm Development (CGP & CGM-0)
- Continuous Glucose Prediction (CGP): Architected a sophisticated LSTM-based RNN to forecast 2-hour post-meal glucose trajectories.
- Performance: Achieved 9.8% MAPE and 13.2 mg/dL RMSE on a held-out test set of 296 users.
- Methodology: Integrated multimodal inputs (demographics, HR, nutrients) and optimized using a meta-learning framework to allow rapid adaptation to individual physiology.
- CGM-0 (Zero-Shot Learning): Engineered a breakthrough glucose prediction model for users who have never worn a sensor.
- Training: Utilized targeted data masking to force the model to learn generalizable relationships between lifestyle and glycemic response without historical glucose data.
- Impact: Achieved 16.6% MAPE, democratizing access to metabolic monitoring by removing invasive hardware requirements.
- Delayed Input Robustness: Solved real-world data gap challenges by training with Gaussian simulated delays ($\mu=2h, \sigma=8h$), enabling robust predictions even with intermittent user scanning behavior.
- Virtual CGM (VCGM) & Digital Twins
- Virtual CGM Engine: Developed a novel estimator for real-time and historical glucose values for post-sensor users.
- Accuracy: Achieved 0.83 Correlation and 13% MAPE by leveraging historical personalized patterns without any recent sensor leakage.
- Enterprise Digital Twins: Designed Digital Twin simulations for Nestle, Mars, and Campbell’s, enabling partners to virtually test food products across diverse user cohorts.
- CES Innovation Award: Implemented activity counterfactuals and alcohol impact analysis, pivotal in winning the CES Innovation Award 2025.
- Research & Leadership
- Nature Digital Medicine: Lead Data Scientist for the "Season of Me" paper (2023) and subsequent 2025 publication. Managed all data analysis and statistical validation for high-impact peer-reviewed research.
- RAG Architecture: Architected a Retrieval-Augmented Generation pipeline ingesting 50,000+ clinical articles to power the "Mirror" health AI.
- Patent Portfolio: Key contributor to the CGP Patent (704.302) and led filings for Food Detect and Displaying Food Activities.
ML Intern
January AI
Menlo Park
10.2020 - 12.2020
- Implemented performance evaluation metrics to assess model efficiency and reliability.
- Conducted comprehensive data preprocessing and feature engineering for model training.
- Collaborated with cross-functional teams to integrate AI solutions into existing workflows.
Education
Certificate - Data Analytics and Visualization Bootcamp
University of California, Irvine
Irvine, CA01-2020
Master of Science - Mechanical and Aerospace Engineering
University of California, Irvine
Irvine, CA01-2019
Bachelor of Science - Mechanical Engineering
University of California, Riverside
Riverside, CA01-2017
Skills
- Machine Learning: Zero-Shot Learning (CGM-0), Meta-Learning, RNNs/LSTMs, Time-Series Forecasting, RAG, Computer Vision (YOLO)
- Data Science: Clinical Benchmarking (MIMIC), Metric Optimization (RMSE/MAPE), Data Augmentation (Gaussian Delays), Hypothesis Testing
- Engineering: Python, TensorFlow, PyTorch, SQL, REST APIs, Docker, AWS, iOS SDKs
- Domain Expertise: Metabolic Health, Glucose Prediction (CGM), Digital Twins, Clinical Trial Analysis
- Machine learning
- Natural language processing
- Feature engineering
- Model development
Timeline
Senior Machine Learning Engineer / Data Scientist
January AI
12.2020 - 01.2026
ML Intern
January AI
10.2020 - 12.2020
Certificate - Data Analytics and Visualization Bootcamp
University of California, Irvine
Master of Science - Mechanical and Aerospace Engineering
University of California, Irvine
Bachelor of Science - Mechanical Engineering
University of California, Riverside