Vishal Seelam

AI ENGINEER

JULY 2024 - PRESENT

VACCINITY AI - FORT WORTH

• Architected and deployed an LLM framework for predicting patient eligibility in clinical trials by analyzing patient notes against specific criteria, achieving 87.3% criterion-level accuracy. Utilized hybrid retrieval methods (BM25, FAISS, MedCPT with CUDA).
• Developed a multi-stage architecture involving:
  • Annotation for extracting relevant sentences and making predictions at the criterion level,
  • Aggregation to compute trial-level eligibility and relevance scores, and
  • Applications for ranking and excluding trials based on patient eligibility, significantly improving clinical trial matching efficiency by 32.6% to 57.2% over competing models.
• Created and integrated an AI-powered voice call bot, which automates on-call patient questionnaires, enhancing patient outreach and recruitment by dynamically matching patients to relevant clinical trials in real time.
• As the sole AI/ML engineer and founding team member, led the end-to-end model development, driving VaccinityAI’s Series A funding efforts by addressing the critical challenge of patient recruitment for clinical trials.

DATA SCIENCE INTERN

MAY 2023 - APRIL 2024

VIInc - Markham, Ontario, Canada (Remote)

• Developed and deployed advanced machine learning models using TensorFlow and BERT to automate real-time categorization and entity recognition for transaction data extracted from NFC-enabled paperless receipts, enhancing customer insights and reporting accuracy.
• Developed an OCR and NLP pipeline, integrating Tesseract for text extraction from scanned receipts and spaCy for efficient preprocessing, followed by BERT transformer for context-aware analysis, leading to precise feature extraction, transaction categorization, and entity recognition.
• Conducted extensive feature engineering to optimize model inputs, utilizing SQL for ETL processes, and leveraging Apache Spark to process and analyze large-scale transaction datasets, enabling real-time analytics and driving actionable business insights through predictive modeling and customer segmentation.

WEB DEVELOPER INTERN

MAY 2022 - AUGUST 2022

APERIGON LABS PVT LTD. - Bengaluru, India

• Designed and Implemented a dynamic, user-friendly website, showcasing the product flashNtap.
• Integrated Google Analytics and custom tracking scripts to monitor page performance and implemented SEO best practices boosting the page’s visibility in search engine rankings to improve user engagement and retention.

B.S. Data Science and Mathematics

2020 - 2025

Texas Christian University, Fort Worth, TX

COURSEWORK

DEEP LEARNING

ARTIFICIAL INTELLIGENCE

SOFTWARE ENGINEERING

ANALYSIS OF ALGORITHMS

DATABASE SYSTEMS

DATA STRUCTURES

DATA SCIENCE

WEB TECHNOLOGIES

UNIX/LINUX

STATISTICS

PROBABILITY

LINEAR ALGEBRA

CALCULUS II

ELECTROCARDIOGRAPHY RISK ENGINE

• Developed an advanced ECG image classification system using a Vision Transformer (ViT) architecture, achieving a breakthrough accuracy of 99% in both diagnosing and classifying raw ECG images without any pre-processing, significantly surpassing existing models' accuracy of 96%.
• Integrated AWS Bedrock and the Anthropic Cloud SONNET model to enhance diagnostic capabilities, enabling the system to classify ECG images and generate precise diagnostic decisions and justifications from a large language model (LLM).

SALIFORT MOTORS EMPLOYEE RETENTION

Developed a data-driven project for Salifort Motors' HR department, focusing on analyzing datasets and building predictive models to forecast employee retention.

WAZE USER CHURN ANALYSIS

a project on user churn analysis at Waze, employing Python for data preparation and conducting in-depth exploratory data analysis (EDA) to identify key user behavior insights.

FROG SCHEDULER

A full-stack web application that allows users to schedule private events for TCU mascot, cheerleaders, and showgirls.