AEROSPACE · NEW DELHI
Aevantis Aerospace
Production marketing site for an unmanned-systems manufacturer, presenting product lines, capabilities, and an enquiry flow.
I maintain the site’s content, hosting, and performance.
Visit siteHi, I'm Pranav Dhawan
Currently building AI workplace automation at the American Chemical Society in Washington, DC. MS in Data Science from George Washington University.
ABOUT ME
I’m an AI Engineer and Data Scientist working on production AI, ML, and analytics systems. I hold a Master’s in Data Science from George Washington University and a Bachelor’s in Computer Science and Engineering. My experience spans computer vision, natural language processing (NLP), and analytics, with a focus on transforming messy data into reliable, scalable solutions.
I’ve recently been working on AI agents, document intelligence, and workflow automation. I’m drawn to problems that involve navigating ambiguity, optimizing under constraints, and building systems people actually use.
I’m always open to connecting with others working on interesting problems, so feel free to reach out.
Experience with
- American Chemical Society
- EY
- HCLTech
- Lumina Datamatics
- Siva Info LLC
- Aevantis Aerospace
- Admiles Media
SKILLS & TECHNOLOGIES
Language
ML
Data & Cloud
Visualization
AI
PROFESSIONAL EXPERIENCE
May 2026 - Present
Current
AI Workplace Engineer Intern
American Chemical Society | Washington, DC
Owned the architecture and production deployment of an AI-orchestrated service-desk workflow integrating Microsoft Copilot, Ivanti, and SharePoint; automated classification, response generation, routing, and follow-up for approximately 245 of 350 monthly tickets.
Engineered an onboarding and offboarding agent that automates lifecycle-aware tracking and documentation of employee hardware assignments, strengthening asset traceability from provisioning through return.
Authored the AI governance framework for formal organizational adoption, defining data-protection, knowledge-access, human-review, and responsible-deployment controls for enterprise AI workflows.
Technologies Used:
June 2026 - Present
Current
Independent Contractor - AI
Siva Info LLC | New York, NY
Owned development of edition-aware PDF-diff tooling that identifies text movement, grammatical edits, deletions, and additions, creating a reliable change-detection layer for downstream content updates.
Re-engineered an XML tagging and classification pipeline across approximately 3,000 entries, increasing completeness and accuracy from 32% to 99% while cutting processing time by approximately 50%.
Technologies Used:
August 2024 - May 2026
Graduated
Master of Science, Data Science
George Washington University | Washington, DC
GPA: 3.74/4.0
Completed a published research paper and technical report from the Multimodal Techniques for Financial Time Series Forecasting capstone project.
Completed dataset reports analyzing U.S. government health, CBP.gov, and WMATA datasets.
Relevant Coursework:
February 2024 - August 2024
Machine Learning Engineer
Lumina Datamatics | Chennai, India
Owned the end-to-end development and AWS SageMaker deployment of a YOLOv8 document intelligence pipeline processing more than 10,000 documents per day, eliminating manual processing and improving extraction accuracy by 16%.
Replaced legacy LayoutParser components with a custom computer vision pipeline evaluated using mAP@50 and mAP@50-95, reducing inference latency by approximately 0.3 seconds per page.
Architected and deployed a hybrid RAG system for internal legal document search using Sentence Transformers, FAISS, BERT QA, and BART, reducing retrieval time from minutes to under 5 seconds.
Redesigned factual extraction around extractive BERT QA after identifying hallucinated structured fields from generative BART outputs, improving grounding for dates, names, and case metadata.
Technologies Used:
August 2020 - May 2024
Graduated
B.Tech in Computer Science and Engineering
Manipal University Jaipur | Rajasthan, India
GPA: 8.51/10.0
Wrote a comparative analysis paper on transformer models versus LSTM for sentiment classification across the X platform.
Relevant Coursework:
July 2023 - September 2023
Machine Learning Intern
HCLTech | Noida, India
Built clustering, Logistic Regression, and Random Forest models across data from 500+ employees, achieving 87% predictive accuracy and identifying the five strongest drivers of workforce performance.
Engineered 12+ behavioral and productivity features using SQL and Python and surfaced model findings through Tableau dashboards.
Technologies Used:
May 2023 - July 2023
Summer Intern
EY | Gurugram, India
Developed Power BI dashboards for Sales and HR functions, visualizing revenue trends, customer acquisition, headcount, attrition, and compensation benchmarks.
Automated ETL workflows across more than five data sources using Alteryx, eliminating manual cleaning errors and supporting 100% reporting accuracy across monthly business reviews.
Technologies Used:
August 2021 - September 2021
Full Stack Intern
LEARNOVATE ECOMMERCE
Developed responsive web interfaces using HTML, CSS, and JavaScript, improving usability across customer-facing digital pages.
Collaborated on website enhancements to improve visual consistency, page structure, and user experience.
Technologies Used:
June 2021 - August 2021
Front-End Intern
Education 4 ol
Built reusable frontend components using HTML, CSS, and JavaScript to improve development efficiency across multiple web projects.
Supported frontend implementation tasks focused on clean structure, maintainability, and design fidelity.
Technologies Used:
CLIENT WORK
Alongside my ML work, I design, ship, and maintain production websites for businesses — the same ownership and operating discipline I bring to AI systems.
OUTDOOR ADVERTISING · DELHI NCR
Admiles Media
Production site for an outdoor advertising company, with media-location information, project galleries, and lead capture.
I maintain the site’s content, hosting, and performance.
Visit siteFEATURED PROJECTS
Stock Screen
A real-time AI-powered terminal that unifies stock analysis, market news, company filings, forecasting workflows, and market context into a single interface. Designed for investors, it helps compare signals quickly and turn live market data into actionable research.
GCP
Python
ML
Multimodal Techniques for Equity Forecasting
Conducted a benchmark study on the FinMultiTime dataset using multiple data modalities, evaluating 10+ different model architectures and ensembles. The findings were published in a research paper and technical report.
Python
LSTM
FinBERT
Attention
Edge-Based PII Detection & Censoring System
Achieved 98.1% F1-score and 97.9% recall across 54 PII entity types by architecting and benchmarking 4 transformer architectures, selecting DeBERTa as the production model and deploying via ONNX for <100ms inference.
DeBERTa
ONNX
Streamlit
Singularity — Local-First Memory for AI Assistants
A local memory and permission layer for AI assistants: it imports your chat history, answers from your own files with citations you can open to the exact source range, and holds every outbound call behind a redaction gate you approve before anything leaves the machine.
TypeScript
RAG
SQLite FTS5
Ollama