cv

Doctoral Researcher

Centre for Addiction and Mental Health (CAMH) - Digital Mental Health Lab

Doctoral research on sycophancy and safety in large language models used for mental health support.

• LLM evaluation
• AI safety
• Digital mental health
• Psychometrics and IRT

Graduate Student Researcher, Biomedical Informatics

Arizona State University - SHARES Lab

Built and evaluated privacy-preserving clinical decision support tools for sensitive health data as part of a NIDA-backed research initiative.

• Built an auditable FHIR and CQL consent engine for granular data sharing under 42 CFR Part 2
• Benchmarked 11,519 synthetic patient records at about 18,900 FHIR resources per second
• Co-authored an Applied Clinical Informatics paper and an AMIA 2025 systems demonstration
• Worked with clinicians, informaticians, and policy experts to turn research prototypes into usable tools

Software Engineer, Investment Banking Technology

Credit Suisse Services AG

Built and supported global order management and compliance systems for stocks, futures, and options across EMEA and US markets.

• Software Development
• C#
• Order Management Systems
• Database Management

Technical Analyst, Investment Banking

Credit Suisse Business Analytics

Designed and improved daily C# ETL pipelines for global trading data, cutting processing time by 70% and making new jobs faster to deliver.

• Data Analysis
• ETL
• SQL
• C#

Technology Intern, International Wealth Management

Credit Suisse

Worked on CI and DevOps improvements for International Wealth Management, including load balancer and OpenShift cluster deployments.

• OpenShift
• CI/CD
• Load Balancers

Undergraduate Researcher, Computational Psychiatry

Visvesvaraya National Institute of Technology, Nagpur

Developed a multimodal depression-detection pipeline using smartphone, audio, and visual signals. Improved data-collection memory efficiency by 90% and co-authored two peer-reviewed papers.

• Quantitative Research
• Multimodal Machine Learning
• Digital Healthcare
• Python

Software Developer, Google Summer of Code 2019

The Apache Software Foundation

Implemented and tested pseudorandom number generators for Apache Commons RNG using Java, Maven, JUnit, and JMH.

• PRNG
• Java

Internship Trainee, Central Technical Services

Reliance Infrastructure (now Adani)

Worked with the central technology team to streamline web application development using JavaScript and the .NET Framework.

• JavaScript
• .NET Framework
• Web Development

Member, AI in Public Health Working Group

NYU x Humane Intelligence

Contributing to interdisciplinary discussions on responsible AI in public health.

• Public Health
• Responsible AI
• Working Group

Peer Reviewer

International Journal of Human-Computer Studies (Elsevier)

Reviewed manuscripts for IJHCS.

• Peer Review

Peer Reviewer

NeurIPS Workshop on Time Series for Health (TS4H)

Reviewed submissions for TS4H workshop.

• Peer Review

Elemental Member

American Association for the Advancement of Science (AAAS) - Neuroscience

AAAS membership in Neuroscience section.

• Membership

Student Member

American Medical Informatics Association (AMIA)

Professional student membership in AMIA.

• Membership

Subgroup Lead, AMIA Student Working Group

American Medical Informatics Association (AMIA)

Led the Trust, Transparency and Justice in Data Use subgroup for the AMIA 2025 Year in Review keynote, reviewing and grouping more than 200 articles.

• Literature Review
• Research Synthesis
• AMIA Year in Review

Career Guidance Outreach

Antarang Foundation

Led career guidance initiatives for at-risk youth in technology careers.

• Outreach
• Education

STEM Mentorship & Diversity Advocacy

Credit Suisse

Co-founded a mentorship program for underrepresented STEM professionals.

• Mentorship
• Diversity
• Community

Technical Advocacy & Content

Credit Suisse

Led engagement for global coding challenges and edited the TA newsletter.

• Content
• Advocacy

Student Tutor

Visvesvaraya National Institute of Technology, Nagpur

Taught computing fundamentals and data structures to underserved students.

• Teaching
• Computing Fundamentals
• Data Structures

Doctor of Philosophy

University of Toronto

Health Systems Research, Institute of Health Policy, Management and Evaluation (Dalla Lana School of Public Health)

• Research focus: sycophancy and safety in LLMs for mental health support
• Based at the CAMH Digital Mental Health Lab
• Canada Impact+ Research Training Doctoral Award (CIRTA)

MS (Thesis)

Arizona State University

Computer Science (Biomedical Informatics)

• Thesis: An Interdisciplinary Framework for High-Throughput, Deterministic Segmentation of Sensitive Health Records in HL7 FHIR
• Herbold Foundation Scholar, 2025

Executive PG Program

International Institute of Information Technology, Bangalore

Data Science

Bachelors (B.Tech)

Visvesvaraya National Institute of Technology, Nagpur

Computer Science and Engineering

• Data Structures and Program Design
• Software Engineering
• Operating Systems
• Neuro-Fuzzy Techniques
• Database Management Systems
• Introduction to Object Oriented Methodology
• Design and Analysis of Algorithms
• Theory of Computation

Canada Impact+ Research Training Doctoral Award (CIRTA)

University of Toronto

Federally funded doctoral award supporting my PhD research at the University of Toronto.

1st Place, ZETIC Melange On-Device AI Hackathon

ZETIC

Won for SumiSense, a privacy-first behavioral health assistant that runs on-device.

Herbold Foundation Scholar

Herbold Foundation

A $10,000 award recognizing academic achievement and research potential in computer science.

Graduate Student Resilience Award

ASU Graduate Student Government

A $500 award recognizing resilience and contributions to the graduate community.

RAVE Awards (2)

Credit Suisse

Recognizing Value and Excellence for partnership, accountability, and project contributions.

Ryan Star Award for Academic Excellence

Ryan Group of Institutions

Awarded for outstanding academic performance.

Psychiatric and Substance-Use History Changes LLM Recommendations in Counterfactual Cases derived from MIMIC-IV Data

Poster, NEURONTO 2026, Toronto

Counterfactual audit of 14 language models on 450 MIMIC-IV emergency cases, testing whether psychiatric and substance-use history changes clinical recommendations.

A privacy-preserving framework for continuous mobile authentication using digital twins and multimodal biometrics

PeerJ Computer Science, 12, e4104

Proposes continuous mobile authentication using digital twins and multiple biometric signals while limiting raw-data exposure.

An Interdisciplinary Framework for High-Throughput, Deterministic Segmentation of Sensitive Health Records in HL7 FHIR

Arizona State University master's thesis, ProQuest publication 32676530

Presents a practical framework for auditable, high-throughput segmentation of sensitive FHIR records.

Assessing the Effectiveness and Scalability of Fast Healthcare Interoperability Resource-Based Granular Data Segmentation Technology

Applied Clinical Informatics, 17(3), 423-433

Evaluates the accuracy, speed, scalability, and cost of SHARES across 11,519 synthetic patient records.

4 Hz, 4 Pages: Just-in-Time Substance Use Relapse Risk Detection from Wearable Time Series Data

NeurIPS 2025 Workshop on Learning from Time Series for Health

Explores just-in-time relapse-risk detection from high-frequency wearable signals.

FHIR-Based Visual Simulation of Consent-Driven Granular Data Segmentation

AMIA Annual Symposium (Systems Demonstration)

Demonstration of SHARES consent-driven segmentation and visualization for privacy-preserving CDS.

FHIR-Based Data Segmentation for Substance Use Disorders: Current Landscape and the Promise of Large Language Models

The Medical Tech Magazine

An invited overview of sensitive-data segmentation, FHIR, and the potential role of language models in substance use care.

Multimodal Depression Detection: Using Fusion Strategies with Smart Phone Usage and Audio-visual Behavior

International Journal on Artificial Intelligence Tools

Studies early and late fusion of smartphone and audio-visual signals for depression detection, reaching 89% accuracy with late fusion.

A novel multi-modal depression detection approach based on mobile crowdsensing and task-based mechanisms

Multimedia Tools and Applications

Combines smartphone, audio, and visual signals in an end-to-end depression-detection pipeline that reached 86% accuracy.

SumiSense - On-Device Behavioral Health Assistant

Privacy-first iOS assistant for daily reflection, longitudinal pattern detection, and anonymized clinician sharing.

• 1st place at the ZETIC Melange On-Device AI Hackathon
• Qwen3-4B
• Chronos
• SwiftUI

SHARES Consent Engine (FHIR + CQL)

Auditable FHIR and CQL consent engine for granular sharing of sensitive health data under 42 CFR Part 2.

• Consent thresholds
• Sensitivity labeling
• CDS Hooks
• HAPI-FHIR integration

FHIRLight: Clinician-Centered FHIR Visualization

In-browser FHIR visualization for patient timelines, clinical risk scores, cohort views, and consent previews.

• TypeScript
• D3.js
• FHIR
• Clinician survey (N=24)

Granular Segmentation Benchmark on Synthetic FHIR Cohort

Evaluation of SHARES across 11,519 synthetic patients, 197.6 million FHIR resources, and 11 confidentiality thresholds.

• 18,909 resources per second
• Under one cent per full run
• Physician-reviewed classifications

HAPI-FHIR Server + CQL Evaluation

HAPI-FHIR main and terminology servers to support Library/$evaluate workflows for consent rules.

• HAPI-FHIR
• CQL
• Terminology

SHARES-CLI

CLI utilities to encode CQL to base64, create FHIR bundles, and POST to HAPI-FHIR endpoints.

• CQL
• FHIR
• CLI

Sharpnr - AI-Powered Academic Assistant

Privacy-first academic assistant that brings Canvas, Slack, and Calendar updates into one planning and study workflow.

• 3rd place at the ASU Agentic AI Hackathon
• FastAPI
• React/Next.js
• MongoDB

MRI Style Transfer with CycleGAN

Generates T2-weighted MRI from T1 scans to augment data and support diagnostic accuracy.

• TensorFlow
• CycleGAN
• Medical Imaging

Indian Sign Language Translator

Open-source Indian Sign Language translator designed for near-real-time use across changing backgrounds and lighting conditions.

• YOLO-v3 hand detection
• Sequence recognition
• Open-source dataset
• More than 100 downloads per year

Genetic Algorithm Implementation

Notebook demonstration of a genetic algorithm solving Travelling Salesman Problem instances with about 50 nodes.

• Fitness functions
• Crossover
• Mutation
• Convergence visualization

Comparison of Clustering Techniques

Compared K-Means, BIRCH, and agglomerative clustering on college data to study how well each method separated public and private institutions.

• K-Means
• BIRCH
• Agglomerative Clustering

Feature Segmentation via Multi-Res-U-Net Masking

Tested MultiResUNet, an architecture developed for medical imaging, on facial-feature segmentation using CelebAMask-HQ.

• MultiResUNet
• Facial Feature Segmentation
• CelebAMaskHQ dataset

Resource Allocation within a Multi-Client Server

Built a multi-client resource manager with wait queues and locking to keep concurrent allocations synchronized.

• Resource Allocation
• Multi-Client Server
• Wait-list Queue

Kerberos Protocol Implementation

Implemented Kerberos-style authentication and ticket generation between clients and services, with encrypted communication channels.

• Kerberos Protocol
• User Authentication
• Key-Dependent Additive Cipher

Command Shell Implementation

Built a Linux shell in C with process control, multiple-command execution, redirection, signal handling, and error reporting.

• OS System Calls
• Linux Commands
• Redirection

Logistic Regression Classifier for Movie Reviews

Built and tuned a logistic-regression sentiment classifier for IMDB reviews using scikit-learn and NLTK.

• Logistic Regression
• IMDB Dataset
• Hyperparameter Tuning