# Prince

**AI Agents ·AI Security · Adversarial Testing · Agent Red Teaming | IIT BHU**

Gurugram, Haryana, India  
(+91) 9565521581 · [princek082@gmail.com](mailto:princek082@gmail.com) · [prince.mst13@iitbhu.ac.in](mailto:prince.mst13@iitbhu.ac.in)  
[LinkedIn](https://www.linkedin.com/in/prince-kumar-5497p) · [GitHub](https://github.com/maniotrix) · [IntentFrame](https://github.com/intentframe/intentframe) · [Portfolio](https://prince-ai-architect.vercel.app/)

---

## Summary

Engineer and founder specializing in **AI security, adversarial testing, and policy-enforced agent systems**. Built [IntentFrame](https://github.com/intentframe/intentframe)—open-source governance and validation for AI execution—while shipping production AI agents and delivering scoped **security assessments** for mobile and backend stacks. Deep hands-on **reverse engineering and adversarial ML** on a private APOS/RSCE codebase (Java client internals, captcha bypass pipelines, reflection-based instrumentation). 8+ years across Python services, Linux environments, and full-stack systems. IIT BHU (B.Tech) and Google Summer of Code alum. Combines an **adversarial mindset** (mapping bypass paths, escalation chains, and failure modes) with clear written reporting for engineering and product stakeholders.

---

## Core Competencies

**AI Red Teaming & Safety** — Prompt injection · jailbreak / policy bypass · multi-turn escalation chains · agent tool misuse · adversarial inputs · LLM/agent robustness testing · abuse-pattern analysis · model-behavior triage

**Agent & Application Security** — Policy enforcement · action validation · resource deny-floors · auth/session review · multi-tenant isolation · API security · IDOR · secret/credential handling · pre-launch readiness reviews

**Offensive & Research Skills** — Reverse engineering · Java reflection / bytecode analysis · captcha & anti-automation bypass · adversarial ML · vulnerability identification · attack-surface mapping · structured test planning · root-cause analysis · prioritized remediation reporting

**Engineering** — Python · Bash · Linux · Docker · AWS · Java · Kotlin · Groovy · TypeScript · automation & test scripting · HTTP client tooling · production system debugging  
---

## Experience

### IntentFrame —  Creator and Technical Cofounder

*January 2026 – Present · Gurugram, India · [github.com/intentframe/intentframe](https://github.com/intentframe/intentframe)*

Open-source framework for **centralized security, governance, and policy enforcement for AI execution**—no agent action reaches files, credentials, network, APIs, or money without validation.

- Architect supervised multi-process runtime (policy registry, executor, server) with deterministic structural constraints and health-monitored service graph  
- Design **action registries**, **resource deny-floors**, and executor packs (filesystem, terminal, network, finance-adjacent bundles)—the same surfaces targeted in agentic red-team scenarios  
- Map adversarial paths continuously: unauthorized tool invocation, harness-level policy bypass, unsafe action composition, and privilege escalation through agent orchestration  
- Ship alpha runtime packages to PyPI (`intentframe-runtime`, `intentframe-supervisor`, `intentframe-native-kit`) for reproducible local deployment and testing  
- Frame product thesis as **prevention over monitoring**: validate before execution, not after incident

### Yong Tech — Autonomous AI research platform with RAG, planning, and tool orchestration (Freelance)

*June 2025 \- December 2025 · New Delhi, India · [https://github.com/maniotrix/KarnAGT](https://github.com/maniotrix/KarnAGT)*

- Built advanced **reasoning agent** for research, problem decomposition, and end-to-end solution delivery  
- Exercised adversarial and edge-case thinking on agent planning loops, tool selection, and output reliability in open-source product work

### Yong Tech — Reverse Engineering, Adversarial ML & Bot Automation (Freelance)

*December 2023 – August 2024 · Private codebase*

Lead developer on **APOS/RSCE** (private) — a Java automation client built on reverse-engineered game internals, anti-automation controls, and ML-assisted captcha solving.

- **Reverse-engineered** obfuscated Java client (`com.rsc.client`, `com.rsc.e`); accessed private fields/methods via reflection (`RSCE_Reflection`) and built a 2,000+ line extension layer (`RSCE_Extension`) over the live applet  
- Designed **sleep-captcha bypass pipeline**: extract captcha sprite from in-memory game graphics → POST to local ML OCR server (`/predict`) → automated keyboard entry — end-to-end evasion of a production anti-bot control  
- Built **OcrUtil** service integration: process-managed OCR server, VM-to-host networking, connection retry/timeouts, captcha dataset collection, and Groovy/Java test harnesses  
- Trained and deployed **neural-network captcha recognition** (Python, TensorFlow) behind the OCR HTTP API; iterated on labeled sample datasets for sleep-word captchas  
- Implemented **PathWalker** pathfinding with **MagicPool** teleport routing, door-state handling, kill switches, and debug telemetry export for scripted automation at scale  
- Stack: Java, Kotlin, Groovy, Python, TensorFlow, Apache Ant, Linux/Windows terminal workflows

### Arc.dev — AI/ML Engineer (Contract)

*October 2022 – July 2023  · Gurugram, India*

- Lead delivery on **AI systems**, Python services, and production integrations for client platforms  
- Own end-to-end technical execution: architecture, implementation, reliability, and production debugging in Linux-based workflows  
- Apply security-minded design on sensitive data flows, auth boundaries, and third-party integrations across full-stack engagements

### Independent Security Consulting — Codementor / Freelance

*October 2016 – Present · Remote*

- **5.0 rating across 93+ reviews** · 7+ years mentoring and delivering scoped engineering and security work  
- Conduct **fixed-fee pre-launch security reviews** for production mobile and backend apps (React Native/Expo, Supabase, Stripe): multi-tenant RLS isolation, auth/session handling, API security, payment integrity, secret scanning, and App Store submission readiness  
- Produce written findings with **Critical / High / Medium / Low** severity, file-level references, and actionable remediation—not generic checklists

### Replyve — Lead Mobile Engineer & Technical Cofounder

*October 2020 – August 2022 · Gurugram, India · Private codebase*

- Shipped real-time messaging platform (Android, XMPP, Firebase) with **AWS-hosted server** and Docker deployment; owned auth, push relay, and production ops

### Google Summer of Code 2015 — P2PSP.org

*May 2015 – August 2015*

**Project:** Integration of GUI, Player and Peer for the P2PSP Python Implementation

- Designed, implemented, and integrated a **graphical user interface** for the Peer in the P2PSP Python implementation  
- Built controls for **peer configuration**, **player** management, and **peer–player communication**  
- Coordinated interaction between peers and player through the GUI for the P2P Straightforward Protocol stack

---

## Selected Projects

| Project | Focus | Link |
| :---- | :---- | :---- |
| **IntentFrame** | AI agent security, policy enforcement, supervised runtime | [github.com/intentframe/intentframe](https://github.com/intentframe/intentframe) |
| **KarnAGT** | Reasoning agents, research automation | [https://github.com/maniotrix/KarnAGT](https://github.com/maniotrix/KarnAGT) |

*APOS/RSCE and Replyve were private codebases; technical scope is summarized under Experience above. Details available on request.*

---

## Education

| Institution | Credential | Years |
| :---- | :---- | :---- |
| **Indian Institute of Technology (BHU), Varanasi** | Integrated B.Tech \+ M.Tech, Materials Science and Technology · Completed 4-year B.Tech coursework (degree unawarded due to early exit from 5-year program) | 2013 – 2018 |

---

## Certifications & Recognition

- **Google Summer of Code 2015** — P2PSP.org · *Integration of GUI, Player and Peer for the P2PSP Python Implementation*  
- **Codementor** — 5.0 ⭐ · 93+ reviews · AI agents & security mentoring since 2016

---

## Languages

- **English** — Professional / bilingual  
- **Hindi** — Native  
- **Spanish** — Elementary *(useful for multilingual prompt and abuse testing)*

---