cv

Computational Scientist

Icahn School of Medicine at Mount Sinai

Leading AI/ML initiatives for precision oncology including multi-agent genomic systems, GPU-accelerated pipelines, and predictive modeling.

• OncoCITE (Submitted to Nature Cancer): Performed systematic EDA on CIViC database (11,312 evidence items, 3,083 publications) quantifying 12 structural bottlenecks. Architected 6-agent solution (Claude Agent SDK, 22 MCP tools) with state serialization and vision-based PDF extraction. Validated on 15-paper corpus: 84% ground truth recovery, 97.8% novel discovery precision, 0% critical errors (n=108). Published at ASH 2025.
• PRIME Model (Blood 2025): Co-developed predictive model for patients receiving BCMA- and GPRC5D-targeting T-cell engagers. Integrated clinical, genomic, and treatment response data. Presented at ASH 2025.
• MMAP Pipeline (56x Speedup): Engineered production pipeline on Minerva HPC using NVIDIA H100 GPUs with Parabricks and RAPIDS. Orchestrated 57 computational processes across 3 integrated workflows. Achieved 95.8% processing time reduction (7 days → 3 hours), enabling 8 patients/day throughput. Enabled same-day molecular tumor board readiness.
• Multi-Omics Integration (Clin Lymph Myel Leuk 2025): Applied modified IntegrAO GNN to MMRF COMPASS cohort (N=655). Achieved 50% classification granularity improvement and 258% high-risk detection enhancement. Identified 18 distinct vulnerability profiles with 94% actionable targets.

Associate Computational Scientist

Icahn School of Medicine at Mount Sinai

Developed production AI systems for clinical decision support, inference optimization, and CAR-T therapy monitoring.

• LLM Inference Optimization: Designed model-routing gateway for cost-performance optimization. Implemented KV cache optimization and prompt compression. Achieved 60-80% inference cost reduction using vLLM batch inference on HPC. Multi-cloud deployment across AWS and Azure OpenAI.
• RAG System (medRxiv, 36 citations): Led development using LangGraph orchestration, BAAI/bge-large-en-v1.5 embeddings, and Mistral-7B. Processed 5,000+ documents achieving 88% clinical effectiveness. First prototype deployed and used by 2-3 clinicians for multiple myeloma research. Presented at IMS 2024.
• CAR-T Adverse Event Prediction (Information MDPI, under review): Led ML development for early CRS detection (N=25 patients). Engineered time-lagged features from wearables and 92-biomarker Olink panel. Achieved 84.62% accuracy (ide-cel), 80.62% (cilta-cel) within 6-hour prediction window. SHAP analysis identified IFN-gamma as cross-product predictor.
• Voice ASR Prototype for Clinical Terminology: Fine-tuned OpenAI Whisper Small for on-device deployment. Trained on recorded patient voice dataset to reduce word error rate for multiple myeloma terminology. Prototype demonstrated improved recognition accuracy for domain-specific medical terms.
• Graduate Student Mentorship: Mentored 6 Carnegie Mellon University graduate students on CAR-T therapy monitoring capstone project. Guided experiment design, baselining, and reporting.

Entrepreneurial Fellow

AYA

Conceptualized and built AYA, leveraging RL-optimized NLP model to extract and formulate questions from academic video content.

• Selected for Summer Sprint at NYU's Entrepreneurial Institute (10 selected from 150 startups)
• Developed end-to-end product combining speech processing, natural language generation, and educational content analysis

Machine Learning Researcher & Teaching Assistant

New York University

Research on AI-generated text detection with IBM Research; taught bioinformatics algorithms.

• AI-Generated Plagiarism Detection (ACL 2023, BEA Workshop): Developed novel multi-faceted NLP approach with Prof. Parijat Dube (IBM Research) achieving 94% accuracy in human-AI text classification. Method employs contrastive loss and LLM-generated paraphrases. [31 citations]
• Teaching Assistant - Data Structures & Algorithms for Bioinformatics: Supported graduate course under Prof. Manpreet S. Katari for 3 semesters, teaching 130+ master's students. Created assignments bridging CS fundamentals with bioinformatics applications.

Technical Lead

Carcrew

Founding technical team member for automotive marketplace startup.

• Built MVP using Django, Flask, PostgreSQL
• Contributed to technical due diligence that secured $2M Series A from TVS Group
• Scaled to 10,000+ DAU

MS (Honors)

New York University, Tandon School of Engineering

Computer Engineering

• Deep Learning
• Machine Learning
• High Performance ML
• Natural Language Processing
• ML for Cyber-Security
• Algorithms & Data Structures for Bioinformatics

BTech

Dr. A.P.J. Abdul Kalam Technical University

Computer Science

ASH 2025 Presentations

American Society of Hematology

Two presentations: OncoCITE multi-agent system and PRIME predictive model

IMS 2024 Presentation

International Myeloma Society

RAG for Multiple Myeloma clinical decision support

Lambda Labs Research Grant

Lambda Labs

GPU compute grant for AI research

NYU Entrepreneurial Institute Summer Sprint

NYU Entrepreneurial Institute

Selected as top 7% (10 of 150 startups) for AYA EdTech startup

NYU Tandon Graduate Scholarship

New York University

Merit-based scholarship award ($8,000/year) for academic excellence

Multi-Omics Integration for Multiple Myeloma Subtypes

Clinical Lymphoma Myeloma Leukemia

IntegrAO GNN on MMRF COMPASS cohort (N=655). 50% classification granularity improvement, 258% high-risk detection enhancement.

PRIME: Predictive Relapse Indicators for Myeloma T-Cell Engagers

Blood

Predictive model for BCMA- and GPRC5D-targeting T-cell engagers. ASH 2025.

OncoCITE: AI-Driven Genomic Evidence Curation

Nature Cancer

Multi-agent system for automated genomic evidence extraction. 6-agent solution with 22 MCP tools, 84% ground truth recovery, 97.8% novel discovery precision. Submitted to Nature Cancer.

A RAG Chatbot for Precision Medicine of Multiple Myeloma

medRxiv

Production RAG system using LangGraph, BAAI/bge-large-en-v1.5 embeddings, and Mistral-7B achieving 88% effectiveness. 36 citations.

Early Detection of CRS Using Wearable Devices Following CAR-T Therapy

Information (MDPI)

ML for early CRS detection (N=25). 84.62% accuracy within 6-hour window. SHAP analysis identified IFN-gamma as predictor. Under review.

Beyond Black Box AI-Generated Plagiarism Detection: From Sentence to Document Level

ACL BEA 2023

Novel multi-faceted NLP approach achieving 94% accuracy in human-AI text classification. IBM Research collaboration. 31 citations.

OncoCITE - Multi-Agent Genomic Evidence Extraction

6-agent system using Claude Agent SDK with 22 MCP tools for automated genomic curation from scientific literature. Submitted to Nature Cancer.

• Systematic EDA on CIViC database (11,312 evidence items, 3,083 publications)
• 84% ground truth recovery, 97.8% novel discovery precision, 0% critical errors
• State serialization enabling pause-resume for long-running extractions
• Identified 24.2% curation errors in expert-curated databases

MMAP - GPU-Accelerated Genomic Pipeline (56x Speedup)

Production pipeline on Minerva HPC using NVIDIA H100 GPUs with Parabricks, RAPIDS, and DeepVariant.

• 95.8% processing time reduction (7 days → 3 hours, 56x speedup)
• 57 computational processes across 3 integrated workflows
• 8 patients/day throughput vs. 0.14 previously
• Enabled same-day molecular tumor board readiness

Voice ASR Prototype for Clinical Terminology

Fine-tuned ASR for multiple myeloma medical terminology recognition.

• Fine-tuned Whisper Small for on-device deployment
• Trained on patient voice recordings
• Reduced word error rate for myeloma terminology
• Improved recognition of domain-specific medical terms

Clinical RAG System

Production RAG system for multiple myeloma clinical decision support. Deployed and used by clinicians.

• LangGraph orchestration with Mistral-7B
• 5,000+ documents processed
• 88% clinical effectiveness
• 36 citations on medRxiv

Graduate Research Mentor

Icahn School of Medicine at Mount Sinai

Mentored 6 Carnegie Mellon University graduate students on CAR-T therapy monitoring capstone project.

• Guided experiment design, baselining, and reporting for CRS prediction using wearables and cytokines
• Coordinated with clinical team on project scope and deliverables
• Students developed ML models for early adverse event detection

AI/ML Integration Mentor

Icahn School of Medicine at Mount Sinai

Mentored team members on AI/ML integration workflows and best practices.

• Trained clinical research staff on RAG system usage and interpretation
• Developed documentation and tutorials for AI tools in clinical workflows
• Guided junior team members on production ML system development

Teaching Assistant - Data Structures & Algorithms for Bioinformatics

New York University

Graduate-level course under Prof. Manpreet S. Katari for 3 semesters (Spring 2022, Fall 2022, Spring 2023).

• Taught data structures, algorithms, and genomic algorithms to 130+ master's students
• Created assignments bridging computer science fundamentals with bioinformatics applications
• Held office hours and provided one-on-one support for complex algorithmic concepts
• Developed course materials covering sequence alignment, graph algorithms, and dynamic programming for genomics