Hitarth Jain
AI security researcher — prompt injection, and the instruction/data boundary that makes it work
Talent
BengaluruMember since September 22, 2026
PythonGoRustPyTorchtransformerslinear probesactivation steeringAWSblockchaineBPFcoordinated disclosure
·
Hire this person
Send a job offer directly to this candidate
PythonGoRustPyTorchtransformers
About
AI security researcher focused on prompt injection and the instruction/data boundary. Build measurement-based benchmarks with reproducible numbers and published write-ups. Research spans web, firmware, cloud and blockchain security, investigating where systems trust input they shouldn't. Available immediately for remote or Bengaluru positions, graduating Sep 2027.
Experience
Independent Security Researcher — Vulnerability Research & Bug Bounties at HackerOne, Intigriti, Bugcrowd (2025-01 – Present)
- Coolblue — reflected XSS chained with an unauthenticated cookie leak into full account takeover. Critical, €1,337 bounty.
- ARM TF-M — uninitialized Secure-world stack disclosure via the psa_fwu_query firmware-update API. $3,000 bounty.
- Command injection → RCE in a security vendor's VS Code extension UNDER COORDINATED DISCLOSURE — 12.5k installs. POSIX shell-escaping applied correctly, then re-quoted by a Windows regex that left the shell metacharacter outside quoting; triggered by merely opening an untrusted repository. Proven on real Windows Server via CI runners; reproduced by the program's triage team.
- Exodus — unauthenticated write endpoint on a production API. Resolved.
- CometBFT — consensus-layer bug affecting 50+ downstream blockchains; further reports across Cosmos SDK, EVM and IBC.
- PayPal — SSRF via protocol smuggling.
Education
B.Sc. Computer Science in Computer Science – Scaler School of Technology, Bengaluru (2024-08 – 2027-09)
Skills
AI security researcherprompt injection