I came to this work through law, then on‑chain investigation.
I now spend as much time on applied AI, data, and the systems that turn both into something someone can act on.
How I got here
Law taught me to separate a record from a story about the record. On-chain work made that concrete: a transfer either happened or it did not.
Since 2021 I have built crypto analytics, explorers, and tooling that investigators can actually use. That is where the investigation work comes from.
I co-founded c4 Academy, which trains law enforcement officers and investigators to work with blockchain evidence. I also write research and build public systems so the method is visible without exposing private infrastructure.
Hands-on systems work
I build and use AI in real workflows: research, writing, investigations, and daily operations. That includes model selection, local and cloud inference, agents with tools, evaluation, and workflow design with human review.
My doctoral research examines how Indian MSMEs actually use AI. That work is about adoption inside smaller organisations, not about training foundation models.
I do not present myself as a deep-learning scientist or a foundation-model researcher. The public claim is applied implementation, with evidence from systems and writing on this site.
Institutions and sessions
-
GST Bhavan, Pune
Focused session with the CGST Pune II Commissionerate on cryptocurrency and VDA taxation, blockchain forensics, and crypto-crime investigation
-
National Police Academy, Hyderabad
Invited session on crypto crime and blockchain evidence
-
Mumbai Police
Invited session on cryptocurrency investigation
-
Bharati Vidyapeeth (Deemed to be University), Pune
Lecture for a teacher training programme on using AI in education
-
c4 Academy
Training for law enforcement officers and investigators
How I work
The work starts by defining the question, the decision it must support, and the records that can be verified. Computation can be deterministic. AI can assist research and review. A person remains accountable for the conclusion.
- Record: establish the underlying fact.
- Interpret: keep inference separate from the record.
- Test: check whether the evidence discriminates between competing explanations.
- Limit: stop before unsupported claims about identity, control, recovery, or criminality.
Private source code, client material, system access, provider configuration, infrastructure details, and case-specific operational records are not public.
What I am exploring
Current experiments include specialist smaller models, evaluation, local and hybrid inference, and tighter agent workflows with tools and review. I am also experimenting with AI-assisted open-source intelligence workflows. That work is exploratory. It is not a public product and it is not an expert claim.