Open by Default
Research, projects, and documentation are all public. This surface is designed for reading and evaluating work at your own pace — no account, no gate.
System Online
Security research, reverse engineering, and AI alignment — built in public by Rishi Saha.
Explore
Capabilities
What gets built here.
Tools and research focused on making systems more understandable, more secure, and more accountable. Each project ships with documentation and open methodology.
repX
Reverse Engineering Toolkit
- Binary disassembly & static analysis
- Packed/obfuscated sample identification
- Cross-architecture reasoning workflows
- Structured output for audit trails
trnX
Transport Analysis
- Network traffic inspection & replay
- Infrastructure mapping & enumeration
- Protocol-level anomaly detection
- Reporting with reproducible captures
Security Research
Offensive & Defensive
- Vulnerability discovery & disclosure
- AI/LLM adversarial red-teaming
- Threat modeling for real systems
- Published findings & methodology
Process
How this site works
01
Read the Research
Security research and AI alignment / red-teaming write-ups — methodology, evidence, and findings published in the open.
02
Explore the Projects
repX for binary analysis, trnX for network inspection — each ships with documentation and developer references.
03
Follow the Blog
Long-form notes on reverse engineering, systems security, and adversarial evaluation as the work progresses.
04
Get in Touch
Available for security consulting, AI red-teaming, and systems engineering — contact and engagement details on the hire page.
Doctrine
Investigate deeply. Publish carefully. Build what can be audited.
This space is intentionally built for serious work at the intersection of systems, security, and AI alignment. The interface is minimal by design so attention stays on signal: methods, evidence chains, and outcomes that can be defended under scrutiny.
Nothing here is decorative filler. Every published finding carries its provenance — what was observed, how it was verified, and what would falsify it. Precision in language, provenance in data, and reproducibility of results are treated as baseline requirements.
Design Position
The objective is not maximal UI. The objective is maximal clarity under pressure.
If a feature cannot explain itself in one pass, it is either unfinished or unnecessary.
P1
Evidence before narrative
Observations stay separate from assumptions. Claims should be reproducible, linkable, and auditable.
P2
Design for revocation
Access and privileges are temporary by default. Everything degrades safely when trust changes.
P3
Systems over slogans
Useful change comes from process design, clear accountability loops, and measurable improvements.
P4
Public impact mindset
Security and governance choices are evaluated by how they affect people, institutions, and opportunity.
Work
Projects, research, and initiatives currently in progress.
Each of these is documented, open to scrutiny, and designed to produce measurable output.
Entry Notes
Everything here is open. Read it, test it, challenge it.
Research, projects, and documentation are fully public — no account required. The work is meant to be scrutinized, reproduced, and built upon.
Reproducible Methods
Findings ship with methodology. Claims are meant to be checked: write-ups link evidence, tooling, and the steps needed to arrive at the same result.
Responsible Disclosure
Vulnerability work follows coordinated disclosure. What gets published is what can be published safely, with affected parties informed first.
Operating Surface
Built for long-horizon work, not short-horizon noise.
Research and methodology are shared as they mature — findings you can verify, tooling you can run, and write-ups that hold up under scrutiny.
Published Work
Research publications, project documentation, and blog posts — designed to be legible for technical and non-technical readers alike.
Current Focus
AI alignment and adversarial evaluation of language models, reverse engineering with repX, and transport-layer security analysis with trnX.