Rajeshwer Jaswal

Applied research scientist and ML engineer with 10+ years at the intersection of medical physics, computational imaging, and deep learning. Specialized in developing and deploying production-grade segmentation models (PyTorch, TensorFlow, U-Net) and quantitative analysis pipelines across PET, SPECT, MRI, and CT — from OSEM reconstruction and RLT dosimetry to multi-parametric MRI modeling and automated anatomical segmentation. Brings a rare combination of imaging domain depth and ML engineering rigor, with a track record of delivering reproducible, regulatory-defensible workflows for biopharma and clinical research sponsors.

Principal Scientist – ML & Imaging at Perceptive (2023-04 – Present)

• Constructed and validated ML-based image segmentation models (PyTorch, TensorFlow, U-Net architectures) for automated animal hotel splitting, CT bed removal, and anatomical ROI placement in rodent and NHP subjects — directly reducing manual analyst intervention and improving pipeline throughput across active biopharma studies.
• Architected and deployed a quantitative MRI analysis suite covering T1/T2 relaxometry, DWI/DTI/DKI/IVIM diffusion modeling, NODDI, DCE, ASL, MRS, PDFF, and ARIA monitoring, alongside automated brain atlas registration and segmentation — including intracranial tumor modeling via ANTs and ANTsPyNet — supporting multiple concurrent biopharma programs.
• Executed an integrated RLT dosimetry workflow across multiple preclinical and clinical studies, combining TAC fitting and calibrated SPECT quantification with OLINDA organ dose modeling to deliver absorbed dose estimates supporting biodistribution, IND-enabling, and early-phase submissions.
• Developed QA/QC frameworks for ML training and validation datasets, evaluating model performance via custom metric pipelines (Dice, IoU, pixel accuracy) with slice overlays, prediction grids, and training curve diagnostics — ensuring reproducibility and regulatory defensibility across imaging modalities.
• Collaborated with cross-functional teams across lab, clinical ops, and analytics to translate sponsor requirements into automated, scalable imaging analysis solutions.

Scientist & Team Lead – PET/SPECT ML Pipelines at Invicro (2020-04 – 2023-04)

• Owned PET and SPECT analytical strategy for specialized and non-standard studies — translating deep knowledge of radiotracer kinetics, partial volume effects, and scanner sensitivity corrections into ML-assisted workflows that automated previously manual analysis steps and extended the team's analytical capabilities beyond established protocols.
• Engineered a Python-based OSEM reconstruction workflow for NHP PET/SPECT studies, incorporating attenuation and scatter correction, PSF modeling, and systematic tuning of iteration/subset parameters and post-filtering strategies (Gaussian, resolution recovery) — validated against phantom-based accuracy benchmarks to ensure quantitative reliability for downstream biomarker analysis.
• Led development and standardization of quantitative analysis scripts (Python, MATLAB, JavaScript) incorporating physics-informed corrections — including decay correction, injected dose normalization, and SUV/SUVR derivation — ensuring reproducible, harmonized outputs across a team of 6 analysts.
• Managed a team of 6 (3 direct, 3 indirect reports); mentored analysts in ML best practices, image reconstruction fundamentals, and pipeline development, consistently meeting sponsor deadlines.

Senior Analyst – Multimodal Imaging & Analysis at Invicro (2016-08 – 2020-04)

• Led PK/PD and biodistribution analyses across preclinical and clinical studies spanning PET, SPECT, CT, MRI, BLI, CFT, Autoradiography, Ultrasound, and ex-vivo Gamma Counting — applying modality-specific corrections (decay, dead-time, attenuation), dose calibrator cross-calibration, and custom Python NCA pipelines for AUC, Cmax, and half-life derivation from plasma and image-derived TACs.
• Owned the full image analysis pipeline from acquisition through reporting: reconstruction, preprocessing (co-registration, motion and partial volume correction), ROI definition, quantitative feature extraction, and statistical reporting with physics-informed QC at each stage.
• Developed and standardized CFT analysis pipelines for high-resolution 3D biodistribution mapping of fluorescently labeled test articles in preclinical models (Python, JavaScript), incorporating depth-dependent attenuation compensation and intensity normalization for cross-subject comparability.
• Acted as liaison between laboratory and analytical departments, translating wet-lab parameters — injected activity, ex-vivo tissue counts, and calibration standards — into analytically consistent imaging pipeline inputs, improving data integrity and process efficiency.

Research Assistant – Optical Imaging at Martinos Center, MGH (2014-06 – 2017-02)

• Developed MATLAB-based analysis pipelines for non-invasive assessment of cerebral energy metabolism and blood flow using two-photon and OCT microscopy in live preclinical models.
• Performed data acquisition, custom script development, equipment calibration, and routine maintenance of advanced optical imaging systems.

M.Sc. in Biomedical Engineering – Tufts University (2016 – 2018)

B.Sc. in Physics – Pennsylvania State University (2010 – 2014)