MS in Artificial Intelligence · The University of Texas at Austin

2025 – Present · GPA 4.0 / 4.0

• Focus: NLP, Vision-Language Models, Multi-Agent Systems
• Compute: TACC Vista (GH200 nodes)

B.Tech in CSE (AI & Data Science) · Kurukshetra University

2020 – 2024 · GPA 9.0 / 10.0

• 100% Merit Scholarship: Awarded a full tuition waiver for all four years of study based on academic excellence.
• University Rank 1: Gold Medalist (First Year).
• University Rank 2: Silver Medalist (Pre-final Year).

AI Engineer · RediMinds Inc.

May 2024 – Present · Remote (USA-based)

IPOS Ocean GPT - Ocean Policy RAG System (npj Ocean Sustainability, Nature Portfolio)

• Independently designed and implemented the entire PostgreSQL schema on GCP, building a system that indexes over 800,000 documents from 47 sources including OpenAlex, Semantic Scholar, FAO, UNEP, and the International Maritime Organization.
• Built automated ETL pipelines using GCP Cloud Run for large-scale parallel ingestion, deduplication, and vector embedding generation across scientific papers, policy documents, and NGO datasets.
• Engineered a multi-agent RAG framework with a dedicated Reference Checker Agent, achieving a 0% citation hallucination rate across all evaluated queries.
• Developed a real-time News Agent integrating Bing Search API, with iterative prompt engineering to ensure source accuracy and recency.
• Built three distinct evaluation frameworks: Agentic, LLM-as-a-Judge (Grok and Deepseek as judges), and a Reference Quality Evaluator, producing the first comprehensive evaluation suite for the system.
• System was successfully piloted with the Government of Seychelles to accelerate circular economy policy decisions around marine plastic pollution.

Evidence Support System - Medical RAG Pipeline

• Designed and evolved a normalized PostgreSQL schema across multiple iterations, ultimately housing 70,000+ clinical documents across ODG, MD, State, ACOEM, MCG, Cascade, and InterQual guideline types.
• Built specialized ingestion pipelines per guideline type, including PubMed and PMC API integration, OCR-based PDF parsing (benchmarked pytesseract, markitdown, and Gemini OCR), and full-document indexing for state guidelines.
• Reduced retrieval latency to under 1 second through two targeted optimizations: switching from 32-bit to 16-bit halfvec embeddings (scaling dimensionality from 768 to 3072 without performance degradation) and implementing a multi-tier indexing strategy combining HNSW, B-Tree, and GIN indexes.
• Validated system reliability using LLM-based evaluation against Claude and Gemini, achieving 100% reference accuracy and top quality scores across all tested clinical cases.
• Refactored backend from FastAPI to Flask, built production endpoints, and deployed an internal Streamlit testing interface used for company accreditation.
• Migrated the entire backend to the new google-genai SDK, integrating support for gemini-3.1-pro-preview and gemini-3-flash-preview with dynamic global endpoint routing.

IRO Project - Healthcare Credentialing Pipeline (28 US States)

• Conducted field-level credentialing mapping across all 50 US states and delivered a structured comparative report to the team.
• Built end-to-end license and sanctions check pipelines for 28 states, covering NPI API integration, state website scraping with reCAPTCHA handling, license number normalization, and automated credential verification.
• Integrated OIG-LEIE federal sanctions checks and standardized all outputs into a unified schema for automated report generation in PDF and markdown.

Salish Sea Digital Twin - Marine Habitat Simulation

• Developed and deployed a YOLOv8-based fish detection and classification model achieving 94% accuracy, enabling real-time marine species identification inside an Unreal Engine 3D simulation.
• Engineered Physics-Informed Neural Networks (PINNs) for underwater hydrology prediction achieving 97% accuracy, with real-time current and tide data visualized as dynamic 3D flowlines and arrows.
• Integrated live NOAA telemetry data and built real-time pipelines synchronizing environmental metrics with EMF visualizations, tidal turbine dashboards, and wave energy modules.
• Contributed to a published methodology for assessing and monitoring marine benthic and pelagic habitats in the Salish Urban Sea System, demonstrating how AI and digital twins can reduce adverse environmental impacts at tidal energy harvesting sites.
• Deployed the entire backend to GCP, independently navigating the platform and delivering a production system used in active marine research.

Agentic AI Systems

• Evaluated CrewAI, LangGraph, n8n, and custom architectures across multiple projects to establish a unified agentic tech stack at RediMinds.
• Decomposed an employee onboarding and offboarding workflow into 14 automatable components and initiated system design for the full agentic pipeline.

ML Engineer Intern · iNeuron.ai

July 2023 – August 2023

• Built a mushroom toxicity classifier on a balanced dataset of 8,124 instances (3,916 poisonous, 4,208 edible), achieving 100% accuracy, precision, recall, and F1 through rigorous data preprocessing and model selection.
• Deployed as a full-stack web application using Python, Flask, HTML, and CSS.