Boston University

Massachusetts Institute of Technology

Passionate about quantitatively exploring the world

Skills: Artificial Intelligence (AI) · Python (Programming Language) · Extract, Transform, Load (ETL) · Machine Learning · Data Visualization

Stealth Mode

Co-Founder, CTO

- Developed first Pan-Disease model. This model was a Transformer architecture with cleverly implemented encoders and probabilistic decoders to cover a wide variety of clinical measurements (both static in time and dynamic, and of variety of types such as binary, ordinal, continuous, etc). It achieved SOTA performance on many internal performance measure across multiple diseases, model size could scale significantly while maintaining training stability, and trained faster than previous single-disease models. This opened the door to training models on much larger datasets than previously possible.
- Created a method to zero-shot learn clinical measurements by combining domain knowledge (scientific literature about the measurements, known correlations, and metadata) with Large Language Models. For measurements with high correlation with trained on measurements, this technique yielded impressive performance at times on par with models trained with the zero-shot learned measurements.
- Tools: PyTorch, PyTorch Lightning, Nextflow, AWS Parallel Cluster, Neptune

Unlearn.AI

Machine Learning Scientist

- Modeling: Developed a SOTA model for predicting major health events. This model used a co-regional Gaussian Process for multi-input/output time-series smoothing and prediction in conjunction with a non-linear Cox Hazard model for event prediction (implemented in TensorFlow). Created a variety of more standard classification models using Scikit-Learn for a variety of business and customer needs.
- Pipelining: Created and maintained Extraction, Transfer, and Loading (ETL) pipelines to aggregate data across multiple servers per client, and across multiple internal data sources and types (DynamoDB, PostgreSQL). These data aggregations fed into further model training and prediction pipelines that were orchestrated using AWS Batch.
- Infrastructure: Created and maintained computational infrastructure (VPCs, DBs, Compute Clusters, etc.) using Terragrunt and Terraform. The infrastructure met strict security and privacy constraints (met HIPAA requirements and was SOC2 certified).

Myia Labs

Data Scientist

• Designed, simulated, built, calibrated/characterized, & automated numerous instruments used by researchers throughout BU's Photonics Center.
- MEMS for precision straining 2D materials (High force & sample thermal isolation)
- Low ripple (< 0.1˚C) temperature controlled stage utilizing Pt RTD & TEC
- High-pressure vessel (>1200 PSI) with sapphire optical window meeting low weight constraint
- Pico-pipette system capable of depositing droplets 4 um in diameter with micron precision
- Raman microscope with Kohler illumination and ultra-long working distance objectives
• Developed sophisticated techniques for fabricating novel micro-electronic & -mechanical devices.
- Method for combining 2D materials with MEMS overcoming limited adhesion & fragile, non-planar structures
- Sub-micron resolution shadow-mask procedure (Achieves ultra-clean metal deposition on delicate structures)
• Contributed to the field of 2D materials in both theory & experiment.
- Monolayer MoS2 trion photoluminescence temperature dependence and trion effective size
- Phonon polarization and energy response to strain in few-layer hBN & phosphorene
- Strain generated pseudo-magnetic fields in monolayer TMDCs & novel geometries in Graphene
• Provided substantial leadership to research team: Essential contributor to research roadmap and grant applications. Planned and executed laboratory transition to 2D materials.  Managed laboratory safety, equipment maintenance, and new member training.

Doctoral Researcher & NDSEG Fellow

• Quickly came up to speed on sensor technologies, resulting in promotion within the first 6 months.
• Developed key hardware characterization metrics leveraged across touch hardware platforms and created corresponding measurement automation tools delivering high accuracy and repeatability.
• Responsible for cross-functional (EE, ME, Firmware, Reliability and Process) coordination to deliver the ultimate in performance. "Best touch experience in the business, hands down" - Engadget (2012 MacBook Pro)

Apple Inc.

Tech Lead, Mac Touch Products

• Created a system to record Biometric and GPS data implemented with 3 PIC processors on an accelerated time schedule. The design also met constraints to be compatible with existing hardware,
exceeded Bit Error Rate requirements and has been deployed successfully in hundreds of tests.
• Upgraded Data Acquisition System: Increased data rate, made the system extensible from three simultaneous measurements to many and incorporated a feedback system to validate measurements. Characterized the measurement systems to evaluate dither and determine noise and precision.
• Designed, built and programmed 915 MHz wireless USB Device implemented with an Altera FPGA
that handled USB and data demodulation in parallel to eliminate simultaneous interrupt errors.
• Managed EmBand production: Coordinated 13 vendors and Contract Manufacturers. Managed a team of 3 full-time, in-house assemblers. Responsible for documentation and quality control.

EmSense Corporation

Electrical Engineer

- Dedicated myself to an all out effort to make the US Ski Team.
- Traveled all over the world training with some of the finest athletes and coaches.
- Leader of my ski team with multipole sponsorships and ranked amongst the top racers in the country.

Aspen Valley Ski Club

Alpine Ski Racer

- First and only undergraduate student to ever be an intern in the Theoretical Physics Department.
- Learned the basics of Quantum Field Theory.
- Developed a collaborative relationship with mentor, Bogdan Dobrescu.

Fermi National Accelerator Laboratory

Intern in the Theoretical Physics Department

- In 2 months, I learned Perl and a variety of client server protocols.
- Created finite state machines to process entries in server log files produced by the protocols.
- Produced and executed a poster presentation with Q and A delivered to a lab wide audience.
- Delivered a lecture to my research group on Roanoke Doppler Systems.

LANL

Cyberdefender

- Rebuilt a drop tube furnace with limited resources, often finding creative solutions using old parts.
- Created automated data analysis tools reducing analysis time from 1 month to 2 days.
- Built a test environment for calibrating the PAS2000, a device so finicky no one could operate reliably.

Jason W. Christopher