Low-drama security consultant who:
- Pokes at things until they confess broken truths.
- Explores AI + offensive security mashups.
- Audits Web3 / DeFi logic and weird edge paths.
- Builds scrappy tools first, cleans them later (sometimes).
- Is still learning, probably forever. (Yes, thatβs the point.)
Motto (today): βShip curiosity. Archive ego.β
I help find failure modes in systemsβinfra, AI pipelines, smart contracts, or βthat legacy service nobody owns.β I prefer:
- Quiet impact over flashy noise
- Reproducible exploit chains over vague βriskβ
- Questions over certainty
- Sustainable craft over heroics
Currently tinkering with: AI-assisted recon, LLM jailbreak surface mapping, hybrid on/off-chain exploit chaining, and silly meta-automation loops that introspect other automation.
| Thread | What Itβs About | Status | Vibe |
|---|---|---|---|
| LLM Prompt Boundary Drift | Injection & semantic leak chains | Ongoing | βContext is porous.β |
| Tool-Augmented Jailbreaks | Chaining plugin/tool misuse | Prototyping | Controlled chaos |
| Hybrid Attack Surfacing | AI + passive recon co-mapping | Iterating | Notebook swamp |
| DeFi Risk Graphs | Protocol dependency cascade | Building | Graph brain |
| MEV Edge Scenarios | Subtle incentive edgecases | Reviewing | Slow burn |
| AI Phishing Variants | Persona + payload staging | Testing | Slightly cursed |
Not an exhaustive list. Just the stuff Iβm fine admitting I break/learn/build with lately.
βStack lists age quickly; curiosity doesnβt.β β Future me, probably.
Unfold Work / Background
- Senior Security Consultant @ NetSPI (2021βPresent, Remote)
- Associate Consultant @ Aujas Networks (2020β2021, Mumbai)
- Executive Cyber-i @ AGC Networks (2019)
- Freelancer / Bug Bounty (2018β2019)
Education:
- B.Sc. Computer Science β Mithibai College, Mumbai
# Recon sketch (harmless outline)
python3 nemesis/recon/passive_enum.py -d target.tld -o out/target.json
# Smart contract artifact scan
python3 papra/chain/scan_contract.py --rpc $RPC --addr $ADDR --rules rules/depth.yaml
# LLM jailbreak variant harness
python3 promptify/lab/harness.py --model gpt-4o --prompt payloads/jailbreaks/oblique.txtflowchart LR
A[Passive Recon] --> B[AI-Assisted Enum]
B --> C[Surface Graph]
C --> D{Worth Probing?}
D -->|Yes| E[Manual Chain Build]
D -->|No| F[Archive]
E --> G[Exploit PoC]
G --> H[Report / Patch Path]
Ping me with a vuln class, an odd LLM failure mode, or a DeFi edge. Iβll probably overthink it politely.
If you got this far: send me your favorite under-loved CVE. I log the interesting ones. No, you donβt get the spreadsheet. Yet.