Ankush Gupta Email & Phone Number

Washington, Dc, Us

Worldwide, Oo

Project #1 – Project Owner and Principal Engineer for a data platform involving end-to-end handling of genomic data (varying in source, size, content) from ingestion to public database submission using various APIs.• Handled daily technical development by mentoring and delegating tasks to multiple direct reports, managing multiple codebases (releases, issues, discussions, documentation, security), and developing optimal SDLC / software best practices.• Managed testing for new releases (alpha, beta, and A/B phases). Included development of custom CI/CD workflows, enabling a 100x improvement in time/cost by decreasing required development iterations.• Scaled active user base from 0 to 1000+ and managed onboarding processes for groups across all sectors.• Presented at top research conferences (live demos/talks) to legislators, scientists, and data practitioners.• Led weekly meetings with various stakeholders to ensure successful project delivery. Consistently exceeded projected timelines, directly translating to a 50% project budget increase and contract extension.Project #2 – Lead Engineer for a pipeline ecosystem to ingest, process, store, and query health data from multiple states across the United States.• Pipelines spanned multiple platforms/environments and used complex storage/querying patterns.• Improved overall latency by 10x (10+ hours to <1 hour) per run, saving >4000 hours of computing time and an estimated $100,000 in computing costs alone.• Proposed and developed a testing sandbox, enabling a decrease in development time and cost (100x faster iterations) through improved testing and custom subsampling procedures for datasets.• Managed technical tasks for multiple direct reports (including code reviews and mentorship).• Personal contributions led to timely completion of critical tasks that were delayed for several weeks (spanning a few months), leading to timelines being exceeded and contract renewal with client.

Worldwide, Oo

Project #1 – Benchmarking a quantum computing based AI model (LAMBEQ) against classical models(RNNs and transformers) on a variety of NLP tasks: sentiment analysis, email classification, text translation, and product categorization.• Pipeline included cleaning/preparing distinct datasets, training multiple deep learning models, automated generation of reports/visualizations, and performance evaluation across transfer learning vs scratch and zero vs fewshot learning scenarios.• Developed integration with various monitoring/deployment tools (MLOps), including the ability to run across cloud platforms (AWS, Azure, GCP) and NVIDIA DGX through robust containerization.• Developed web scraping and automated workflow orchestration capabilities (with various checks).• Published findings in a comprehensive white paper and presented at top conferences/symposiums.Project #2 – Development of custom deep learning models for computer vision applications within afederated learning framework (biomedical images).• Trained and evaluated performance of multiple neural networks (CNN and transformer based) on biomedical images and developed custom integration with NVIDIA FLARE (federated learning).• Developed a custom subsampling algorithm for individual FL client sites.• Resulted in improved precision by 3%, recall by 4%, F1 score by 3.5%, and an overall 2.5% increase in test accuracy on average vs non-federated base models.

Bethesda, Md, Us

Lead Researcher and Engineer for a major project involving the construction of an end-to-end data pipeline (ingestion, processing, statistical analysis, and Deep Generative AI models), to elucidate dynamical signatures within the brain during visual tasks and apply these insights to enhance existing theoretics around AI algorithms/architectures related to computer vision.• Wrote ~80% of code for a pipeline to process large amounts (>500TB) of highly unstructured data and utilize downstream AI models with custom functions/embedding (RNN, deep generative, reservoir computing).• Reduced total time spent on processing by 5x (5 hours to 1 hour) per session (2-3 per week), per lab member (6-8 members), for a total of >12,000 hours saved. Pipeline was fully parallelized/automated and consisted of motion correction, signal denoising, and spike deconvolution.• Implemented and explored tools from manifold learning, regression, probability/information theory, and physics.• Implemented various MLOps tools for model logging, experiment tracking, and hyperparameter optimization sweeps.• Developed simulations to closely mimic dynamics of neural networks within the brain.• Presented findings at top research conferences (COSYNE, SfN, Brain Criticality Conference).

Baltimore, Md, Us

• Recruited by Johns Hopkins to conduct research on ferromagnetic nanoparticles as therapeutic agents for targeting specific cancer biomarkers, as an alternative or in conjunction with existing treatment (radiation)• Work involved heavy mathematical/physics modeling of the dynamics between radiation and nanoparticle thermodynamical properties on the biological environment (from local to systems level). Worked closely with mathematicians/physicists at the National Institutes of Standards and Technology (NIST) for extracting specific physical properties of the nanoparticles themselves, in order to utilize this information within the modeling/simulations • Handled core managerial and administrative tasks in order to ensure team deadlines are met and progress is properly recorded• Training of new members on the team in both wet bench and computational/mathematical techniques

Bethesda, Md, Us

Recipient of the prestigious and highly competitive Student Intramural Research Training Award (IRTA) (undergraduate) through the National Institutes of Health (NIH). • Understanding the role of KOR mechanisms within specific neural circuits in the pre-frontal cortex (PFC) and nucleus accumbens (NAcc) areas of the brain as it underlies motivation and emotional regulation. The linkage between these relatively fundamental states and bounds associated with neuropsychiatric disorders was constructed via a variety of behavioral paradigms• Involved looking at specific receptors/transmitters as information carriers between neurons at the circuit level and how perturbations to these dynamics via pharmacological manipulation fits into/solidifies this developed framework• Wrote code for automated analysis of images acquired from histology experiments, which included automatic ROI identification of neurons and global adjustment of tissue orientation and separation of poor quality regions

Richmond, Virginia, Us

• Analysis at the molecular, cellular, and organismal levels on the functional role of Shank3 and HCN channel irregularities on common neuropsychiatric symptoms exhibited in Autism Spectrum Disorders (ASD)• Biophysical alterations of HCN channels located in VB and RTN thalamocortical neurons within the thalamus were explored via neural network/circuit simulations, in order to better understand and mimic the specific biophysical changes that cell receptors undergo during pharmacological experiments• Findings culminated to a senior thesis (B.S. with Honors Distinction), a research poster in UROP, and co-authorship in a top neuroscience journal (ACS Chemical Neuroscience):

Ms In Computer Science, Cs / Ds

Graduate Courses @ National Institutes Of Health

Bachelor'S Of Science (B.S.) In Neuroscience & Chemistry