Arghya Mukherjee, PhD

I’m a data scientist and AI engineer who builds systems that turn messy data into decisions. My work spans vision-language models, generative and agentic AI, fraud detection, ontology development, and MCP servers that connect language models to real tools and data.

As a Senior Data Scientist for the State of Oklahoma, I deliver machine learning and AI across state agencies. I also founded ChainDiligence, a blockchain intelligence company that turns my cryptocurrency crime research into products for screening, monitoring, and investigations.

I earned my PhD in Computer Science from the University of Tulsa in December 2025. My research measures scams, market manipulation, and security shocks in cryptocurrency ecosystems, and has been covered by Bloomberg and Bruce Schneier.

Generative & agentic AI

LLM applications, retrieval-augmented generation, multi-step agents, and MCP servers that give models safe access to tools and data.

Vision & multimodal models

Vision-language models and deep learning for images and video, from training and evaluation to production deployment.

Fraud detection, anti-abuse & AML

Models and data products that surface pump-and-dump schemes, scams, and illicit fund flows in financial and blockchain data.

Ontologies & data platforms

Ontology and knowledge modeling, data pipelines, and analytics that let teams across an organization share one view of their data.

1. Founder

   May 2026 – Present

   ChainDiligence, Tulsa, OK

   • Founded a research-driven blockchain intelligence company that helps compliance, fraud, security, and investigation teams act on blockchain activity.
   • Leading a suite of eight connected products in active development, covering wallet screening, transaction monitoring, fraud prevention, and crypto investigations.
   • Turning peer-reviewed research on pump-and-dump schemes, exchange failure, and cryptocurrency scams into explainable risk signals.
   • Designing risk scoring that shows its evidence: labels, fund-flow paths, confidence, and sources behind every decision.

   Projects

   Address screening

   API that screens wallets for sanctions, scam, mixer, and darknet exposure, including indirect risk.

   Transaction monitoring

   Continuous monitoring of deposits, withdrawals, and swaps with configurable alerts.

   Investigation workspace

   Graph tool for following fund flows across hops and chains, saving case evidence, and exporting reports.

   Market-integrity alerts

   Detection of token manipulation and abnormal trading, informed by pump-and-dump research.

   Tools: Blockchain analytics, Graph intelligence, Entity attribution, Risk scoring, APIs & webhooks

2. Computer Vision Engineer

   Oct 2025 – Apr 2026

   Chevron, USA

   • Built a computer vision inspection system that detects damage to energy infrastructure from imagery.
   • Replaced slow manual review with automated detection, so inspectors can focus on the assets that need attention.
   • Took models from data collection and labeling through training, evaluation, and deployment into inspection workflows.

   Tools: Python, PyTorch, Computer vision, Deep learning, Object detection

3. Data Scientist III

   Jan 2025 – Present

   Office of Management and Enterprise Services, State of Oklahoma, Oklahoma, USA

   • Design, build, and deploy machine learning and AI solutions for state agencies, from scoping with stakeholders to production and monitoring.
   • Build vision-language models that interpret infrastructure imagery and answer questions about it in plain language.
   • Led two generations of roadway asset detection, moving from a GIS-based first version to a computer vision system that cuts manual inspection time and cost.
   • Built a custom image viewer so staff can review imagery and model detections side by side.
   • Shape AI projects in digital twins and noise mitigation, from early design through delivery.

   Projects

   Vision-language models

   Multimodal models that describe, classify, and answer questions about infrastructure imagery.

   Digital twin design

   Project design for a digital twin that brings asset, imagery, and sensor data into one model of the network.

   Noise mitigation AI

   Machine learning that helps assess and prioritize noise mitigation work.

   Roadway assets v1

   First release built on ESRI, with dashboards to explore assets by county, road, and asset type.

   Roadway assets v2

   Computer vision rebuild that detects and catalogs roadway assets directly from imagery.

   Custom image viewer

   Purpose-built viewer for browsing imagery with detections overlaid, designed for reviewers.

   Tools: Python, PyTorch, VLMs, Computer vision, ESRI, Digital twins, Cloud ML

4. Data Scientist, ML/AI Engineer

   Apr 2024 – Dec 2024

   Workonward Inc, Remote

   • Designed voice search and an AI chatbot that let customers reach key information hands-free, increasing user interaction by 200% within three months of launch.
   • Designed and implemented the backend for ML features that score résumés against job descriptions.
   • Built a custom recruiter and job-seeker chatbot on an LLM using retrieval-augmented generation.
   • Worked with IT and DevOps to ensure quality assurance and data integrity.

   Tools: Python, LLMs, RAG, LangChain, Amazon Bedrock, NLP

5. AI Software Engineer

   Apr 2024 – Present

   Anti-Phishing Working Group, Remote (part-time)

   • Develop ontologies that give APWG’s eCrime data a shared structure, linking scams, wallets, domains, and campaigns.
   • Build MCP servers that let AI assistants query and reason over APWG data through controlled, auditable tools.
   • Apply language models to classify, enrich, and connect incoming threat reports for fraud, anti-abuse, and AML use cases.

   Projects

   eCrime ontology

   A common model of threat entities and their relationships across APWG datasets.

   MCP servers

   Tool interfaces that connect AI agents to APWG data with access controls.

   Tools: Python, Ontologies, Knowledge graphs, MCP, LLMs, AWS

6. Data Scientist, ML/AI Engineer

   May 2023 – Apr 2024

   Robbie AI, Remote

   • Developed computer vision models to identify body structures, and neural networks that detect pain level from FACS and PSPI scores.
   • Raised accuracy of existing production models by at least 23%.
   • Built a real-time video pipeline with Azure Blob Storage for live feeds and Apache Kafka streaming data to deep learning models.

   Tools: Python, Computer vision, Deep learning, Azure Blob Storage, Apache Kafka

7. Software Engineer (DevOps)

   Oct 2020 – Mar 2024

   Anti-Phishing Working Group, Remote (part-time)

   • Built and ran data services for APWG’s Cryptocurrency Working Group, used by members in academia, industry, and law enforcement-facing research.
   • Maintained a curated database of roughly 1.2 million Bitcoin wallet addresses linked to scams and ransomware, used by members for fraud prevention, anti-abuse, and AML screening.
   • Kept client-facing systems at 99.999% uptime through testing, monitoring, and CI/CD with Git, Jenkins, and Apache Airflow.

   Projects

   Cryptocurrency scam address feed

   Collection, review, and enrichment of wallet addresses reported in scams and ransomware campaigns.

   Member data APIs

   REST APIs that give working-group members programmatic access to eCrime data.

   eCrime data lake

   Snowflake warehouse that turns raw reports into aggregated, query-ready datasets.

   Workflow automation

   Scheduled ingestion and processing pipelines orchestrated with Airflow and deployed through CI/CD on AWS.

   Tools: Python, R, REST APIs, Snowflake, Elasticsearch, Apache Airflow, AWS, Jenkins

8. Research Assistant & Data Scientist

   Jan 2018 – Dec 2024

   The University of Tulsa, Tulsa, OK (research)

   • Built fraud detection research end to end: pump-and-dump schemes, scams, market manipulation, and the failure of coins and exchanges.
   • Measured anti-abuse and anti-money-laundering signals in blockchain data, tracing how illicit funds move through exchanges and services.
   • Developed ETL pipelines for Bitcoin, Ethereum, and Solana data with Python, Snowflake, and SQL, plus web crawlers and API clients for market and scam data.
   • Combined econometric models with neural networks and NLP to detect manipulation and classify large volumes of text.
   • Published peer-reviewed work at ACM CCS workshops, APWG eCrime, HICSS, and Information Processing and Management.

   Projects

   Pump-and-dump detection

   Identifying coordinated manipulation campaigns and measuring their economic impact.

   Exchange risk

   Survival analysis of why cryptocurrency exchanges close and what it costs their users.

   Scam & fraud measurement

   Large-scale datasets of fraudulent projects, wallets, and campaigns for empirical study.

   AML & anti-abuse

   Fund-flow analysis that links illicit activity to the services it passes through.

   Tools: Python, R, Stata, Snowflake, SQL, NLP, Web scraping, Statistics

9. IT Risk Assurance Associate

   Jan 2015 – Dec 2016

   PricewaterhouseCoopers, Birmingham, UK

   • Delivered IT risk advisory work: risk assessments, security reviews, compliance assessments, and IT governance assessments.
   • Evaluated client IT environments for risks and control gaps, and recommended remediation.

   Tools: Risk assessment, IT governance, Compliance, Security reviews

Security has run through my career, from IT risk advisory at PwC to eCrime data at the Anti-Phishing Working Group, a PhD measuring cryptocurrency crime, and now building blockchain intelligence at ChainDiligence.

Cryptocurrency crime measurement

Empirical studies of pump-and-dump schemes, scams, exchange closures, and the lifespan of coins, published at venues including ACM CCS workshops, APWG eCrime, and HICSS.

Blockchain intelligence & investigations

Wallet screening, entity attribution, fund-flow tracing across chains, and explainable risk scoring for compliance and investigation teams.

Fraud, anti-abuse & AML

Detection of pump-and-dump schemes, scams, and illicit fund flows across University of Tulsa research and APWG work, built on labeled data from research and industry feeds.

Phishing & eCrime data

Collecting, enriching, and serving threat data on scam and ransomware wallets for APWG’s global membership.

Security economics

Measuring the impact of breaches and security shocks on markets, and modeling how organizations should invest in security.

IT risk & governance

Risk assessments, security reviews, compliance testing, and control-gap analysis for enterprise clients at PwC.

My research measures fraud and manipulation in cryptocurrency markets. It has 190 citations on Google Scholar (h-index 3) and has been covered by Bloomberg and Bruce Schneier.

Selected publications

1. Beyond the hype: Empirical evaluation of cryptocurrency unicorn success

   Arghya Mukherjee, Tyler Moore

   Hawaii International Conference on System Sciences (HICSS), 2025. Listen to an audio summary

2. Cryptocurrency Exchange Closure Revisited (Again)

   Arghya Mukherjee, Tyler Moore

   APWG Symposium on Electronic Crime Research (eCrime), 2022

3. An examination of the cryptocurrency pump-and-dump ecosystem

   JT Hamrick, Farhang Rouhi, Arghya Mukherjee, Amir Feder, Neil Gandal, Tyler Moore, Marie Vasek

   Information Processing and Management, 2021. Listen to an audio summary

4. Analyzing Target-Based Cryptocurrency Pump and Dump Schemes

   JT Hamrick, Farhang Rouhi, Arghya Mukherjee, Amir Feder, Neil Gandal, Tyler Moore, Marie Vasek

   ACM CCS Workshop on Decentralized Finance and Security, 2021

In the press

• Bloomberg Reported on our study of pump-and-dump schemes in cryptocurrency markets.
• Schneier on Security Security expert Bruce Schneier wrote about the same research on his blog.

AI & machine learning

PyTorch, TensorFlow, scikit-learn, Hugging Face Transformers, Vision-language models, OpenCV

Generative & agentic AI

LLMs, RAG, AI agents, MCP servers, LangChain, Amazon Bedrock, Fine-tuning

Knowledge & data modeling

Ontology development, Knowledge graphs, Data modeling, SQL, GraphQL

Data & analytics

Python, R, Pandas, PySpark, Stata, Power BI, Tableau, ESRI

Cloud & MLOps

AWS, Azure, Google Cloud, Docker, Kubernetes, Apache Airflow, Snowflake, CI/CD

Security & blockchain

Blockchain analysis, Fraud detection, AML, Anti-abuse, Threat intelligence, SIEM, Elasticsearch, IT risk assessment

1. PhD, Computer Science

   2018 – Dec 2025

   The University of Tulsa, Tulsa, OK

   Control + Alt + Deceive: Empirical analysis of cryptocurrency ecosystem and cyber crime

   Advisor: Dr. Tyler Moore

2. MSc

   2014

   University of St Andrews, St Andrews, Scotland

   Measuring enterprise IEEE 802.11 (Wi-Fi) standards

   Advisor: Dr. Colin Allison

3. Bachelor of Computer Applications

   2012

   West Bengal University of Technology, Kolkata, India

   Computer applications and information technology

I write about data science, AI, and research on Substack. Subscribe or browse the archive.

I’m looking for Principal Data Scientist, Forward Deployed Engineer, and security research roles, and I’m glad to talk about research collaborations or anything in data science, AI, and cybersecurity. Email is quickest, or book a time on my calendar.

• Emailarghyamukherjee1991@gmail.com
• Academicarghya-mukherjee@utulsa.edu
• CalendarBook a call
• Based inTulsa, Oklahoma