About Research Experience Projects Skills Contact
Open to Relocation · Bengaluru, India

Debasish
Tripathy

AI/ML Engineer & Security Researcher. Adversarial ML expertise. 4 granted patents. Responsibly disclosed vulnerabilities to Google, Microsoft, IBM, and Mistral AI. Author of PHANTOM — achieving 98.6% data utility with sub-0.1% re-identification rate.

Apple WWDC 2026 MSRC Acknowledged $5K AWS Activate 4 Granted Patents PHANTOM Research · 2025
Get in Touch View Work
4
Granted Patents
5+
Disclosures
10K+
Users Served
$5K
AWS Credits
Google Imagen 3 — Issue #432059218 MSRC AI Safety — Case 103869 Apple WWDC 2026 Swift Student Challenge AWS Activate — $5,000 Credits PHANTOM — 98.6% Data Utility Pine Labs Hackathon — 3rd Place 4 Granted Patents 2024–2025 IBM Granite · HackerOne #3438834 Google Imagen 3 — Issue #432059218 MSRC AI Safety — Case 103869 Apple WWDC 2026 Swift Student Challenge AWS Activate — $5,000 Credits PHANTOM — 98.6% Data Utility Pine Labs Hackathon — 3rd Place 4 Granted Patents 2024–2025 IBM Granite · HackerOne #3438834
01 / About

Building safe intelligence
that matters.

I'm an AI/ML engineer and security researcher focused on adversarial machine learning, responsible AI, and LLM security. I've responsibly disclosed vulnerabilities in Google Imagen 3, received a Microsoft AI Safety Acknowledgment (MSRC Case 103869), and reported safety issues to IBM and Mistral AI.

I author research — PHANTOM achieves 98.6% data utility with sub-0.1% re-identification rate across 6 attack types. I ship production ML systems: RAG pipelines, on-device inference, real-time anomaly detection.

debasishtripathy2021@gmail.com linkedin.com/in/debasishtripathy48441b1b5 github.com/debasishtripathy13 +91 88970 08513

Core Strengths

AI/ML Engineering
95%
Cybersecurity
92%
LLM Security
93%
Computer Vision
90%
Cloud & MLOps
88%
02 / Security Research

Trusted by the world's
best AI teams.

AI Safety Vulnerability — Microsoft AI System
Reported confirmed vulnerability. Fix released. Acknowledged on Microsoft's public AI Safety Acknowledgments Page.
MSRC Case 103869 · Mar 2026
Fix Released
Prompt Injection — Google Imagen 3
Discovered adversarial prompt injection enabling manipulation of image generation outputs. Contributed to production safety release.
Issue #432059218 · Aug 2024
Acknowledged
Base64 Prompt Injection — IBM Granite 4.0
Base64-encoded prompt injection leading to safety bypass and malicious code generation. Passed to Granite team.
HackerOne #3438834
Informative
Safety Vulnerability — Mistral AI Model
Acknowledged and resolved by Mistral AI Security team.
Ticket #10052633 · Dec 2025
Resolved
Privacy & Data Leakage — Google Gemini
User images exposed via third-party links posing identity risks. Reviewed by Trust & Safety.
Issue #432059218 · Jul 2025
Reviewed
03 / Experience

From research to
production.

IT/ISC Security & Risk Compliance Intern
Philips Healthcare
Jul 2024 – Mar 2026 · Part-time · Bengaluru
  • Established the AI Center of Excellence (AI CoE) to standardize enterprise-wide AI governance and guidelines.
  • Spearheaded end-to-end cross-functional automation and deployed an Automated SDLAN Security Check & Approval System.
  • Engineered a Certificate Lifecycle Management (CLM) platform via ADCS, .NET, Python, and AWS, achieving 100% uptime with zero-downtime.
  • Built PKI-based ML model certificate verification across 50+ FDA/CE-regulated medical device types.
AI CoEAutomationSDLANCLM.NETPythonAWSPKI
Machine Learning Engineer
MedDBot
Jun 2025 – Jan 2026 · Remote · Contract
  • LLaMA 3.1-8B offline medical prescription: 97% accuracy, sub-500ms CPU latency, zero data exfiltration.
  • Real-time anomaly detection for 200+ IoT devices: 94% precision with Isolation Forest + LSTM.
  • Reduced inference cost 4× via INT8 quantisation maintaining 98.5% output quality.
LLaMA 3.1INT8LSTMEdge ML
AI Safety Researcher — Trusted Tester
Google Labs
Aug 2024 – Jan 2025 · Remote
  • Prompt injection in Imagen 3 — contributed to production safety release.
  • Automated adversarial testing: 500+ attack vectors, 8 unknown failure modes.
  • Open-source evaluation toolkit with 200+ GitHub stars.
Adversarial MLRed TeamingGenerative AI
Machine Learning Engineer
Genie AI, Inc.
Apr 2024 – Jun 2025 · Ontario, CA (Remote)
  • Healthcare chatbot with RAG: 89% query satisfaction across 5K+ conversations.
  • Thermal fire detection on Raspberry Pi 4: 95% accuracy at 15 FPS.
  • Reduced deployment footprint 75% via structured pruning.
RAGEfficientNetTFLiteEdge ML
ML Research Intern — Firefly Team
Adobe Inc.
Nov 2023 – Dec 2023 · Bengaluru
  • Curated 100K+ image-text pairs; CLIP filtering improved semantic alignment by 23%.
  • Automated quality assessment — reduced manual curation by 60%.
Adobe FireflyCLIPPyTorch
04 / Projects

Systems built for
real impact.

Security · 2025
Project Libra
Multi-Agent Security Operations Platform
Multi-agent LLM pipeline for automated security log analysis. Hash-chain audit trail with 100% tamper-detection. Isolation Forest + LSTM anomaly detection at 92% precision / 88% recall.
LLM OrchestrationAnomaly DetectionAWS Lambda
12min MTTD
from 4 hours
500K+logs/day
# MTTD pipeline
from libra import AgentPipeline
pipe = AgentPipeline(
  model="gpt-4o",
  detect="LSTM+IF")
Fintech · 3rd Place Pine Labs
SentinelPay
AI Payment Governance for Autonomous Agents
Payment security layer for agentic AI: ML anomaly detection, SHA-256 hash-chain audit, Gemini 2.0 Flash threat analysis, Redis pub/sub kill switch.
Gemini 2.0RedisPine Labs API
SHA-256
Cryptographic audit chain
# Kill switch
redis.publish("sentinel",
  "HALT_AGENT_TX")
# Threat analysis
score = gemini.analyze(tx)
Apple WWDC 2026
StellaPaw
On-Device Pet Health AI · Swift Student Challenge
iOS app using Apple's on-device Foundation Models API for pet health diagnosis, symptom tracking, and vet summaries — zero data exfiltration by design.
Foundation ModelsCoreMLSwiftData
0bytes exfiltrated
// On-device inference
let model = Foundation
  .onDevice()
let diag = await
  model.diagnose(symptoms)
Cryptography · 2025
DIT Protocol
Decentralised Identity — CA-free TLS Alternative
Post-quantum identity protocol using ML-DSA/ML-KEM, FROST threshold signatures, and BLAKE3 Merkle transparency log. Formal security proofs + performance benchmarks.
Post-QuantumFROSTFormal Proofs
PQ-Safe
Post-quantum identity layer
# BLAKE3 Merkle log
log = MerkleTree(BLAKE3)
id = DIT.register(
  algo="ML-DSA",
  sig="FROST")
05 / Research & IP

Publications &
Patents.

Published · 2025
PHANTOM — Privacy-Hardened PII Protection Framework
Debasish Tripathy · Dr. Gururaj · Dr. Shreyas
Neural semantic projection using ViT + LLMs + locality-sensitive hashing. Achieves 98.6% data utility with sub-0.1% re-identification rate against 6 attack types.
NMI 2025
Comparative Analysis of Supervised Learning Evaluation Metrics
Debasish Tripathy · Dr. Deepti Sisodia
12 metrics across 8 ML domains. Domain-aware model selection framework.
IJAC 2024
Vastu-Based Architectural Assessment Using Computer Vision
Debasish Tripathy · Dr. Tusar
CV system analyzing floor plans for Vastu Shastra compliance.
IN 202541045118
Vastumeter — Vastu Analysis System
IN 202441086519
Fashion Assessment System — AI/CV
IN 202541087250
Smart Neck Accessory IoT Device
©37205/2024
Cipher Decryption Algorithm
06 / Technical Stack

The full toolkit.

Languages
PythonC++SQLBashJavaScriptSwift
AI / ML
PyTorchTensorFlowJAXHugging FaceONNXCoreML
Cloud
SageMakerLambdaAzure MLDockerKubernetesMLflow
GenAI & LLM
RAGLangChainRed TeamingPrompt InjectionAdversarial ML
Data
PostgreSQLRedisPineconeWeaviateFastAPI
Responsible AI
SHAP/LIMEBias AuditingFDA/CEAI GovernancePKI
07 / Recognition

Built to win.

🥇
Schneider Electric Cybersecurity CTF
1st Place · 2024
🥇
Coder Army System Design
1st Place · 2024
🏆
Autodesk EduHack
Winner · 2024
🏆
Microsoft AI Odyssey India
Winner · 2024
🥈
PwC Firecrest Hackathon
2nd Place · 2025
🥉
Pine Labs AWS Playground AI
3rd Place · SentinelPay · 2026
08 / Education

Academic foundation.

B.Tech, Computer Science & Engineering
Manipal Institute of Technology, Bengaluru
Specialisation: Cybersecurity · Class of 2026
8.22
CGPA / 10.0
Certifications
Oracle Cloud Data Science Professional
ID: 102993280
Nov 2025
Oracle Cloud AI Foundations Associate
Oct 2025
Cisco CCNA Enterprise Security
Jun 2025
AWS Solutions Architect (SAA-C03)
In Progress
Available · Open to Relocation

Let's build
something
important.

Interested in responsible AI, LLM security, adversarial ML, or collaboration? Reach out — I respond promptly.

Send Email Resume
Email
debasishtripathy2021@gmail.com
LinkedIn
debasishtripathy48441b1b5
GitHub
debasishtripathy13
Microsoft MSRC
AI Safety — Case 103869