Systems builder working across AI security, scientific computing, quantitative finance, and developer infrastructure.

I build systems at the seams between fields. I treat AI agents, financial markets, physical processes, and programming languages as the same kind of object: a thing with state, dynamics, failure modes, and an interface, all of which can be modeled, instrumented, and broken on purpose.

My public work spans:

• AI security — aastf, an Agentic AI Security Testing Framework published on PyPI that intercepts the agent execution graph and tests it against the OWASP Top 10 for Agentic Applications 2026.
• Scientific computing — streaming NASA Kepler-mission data to classify habitability candidates, and a space-mission dashboard.
• Language and compiler design — adarshLang, a small interpreted language with its own lexer, parser, and evaluator.
• Algorithmic depth — a 30-week, four-language curriculum (Java / Python / C++ / Rust) covering everything from binary search to network flow, NP-completeness, and computational geometry.
• Quantitative finance — research and tooling around implied-volatility surfaces, SVI / SABR calibration, HMM regime detection, and Monte Carlo VaR. (Most of this work lives outside this GitHub account.)

If a problem requires translating between physics, finance, machine learning, and production code, that is the kind of problem I want.

• MBA candidate, IIM Raipur — PGP, Class of 2027.
• BSc (Hons.) Physics — undergraduate background in classical mechanics, statistical physics, and computational methods.
• Published author — Realizing a Fully Functional CPU Using Multi-Layer Perceptrons, IJARIIT, Vol. 11, Issue 1, pp. 248–256 (21 Feb 2025), ISSN 2454-132X. The paper introduces IC 616 Ultra-MLP, a fully perceptron-based CPU showing that arrays of threshold units, with appropriate weights and biases, can theoretically implement all standard computing tasks — pipeline, memory hierarchy, and on-chip training included. It sits at the boundary of neural-inspired computing and computer architecture.
• Co-founder, Cosminder Solutions (Deoghar, Jharkhand) — AI research lab. Led work on large-parameter simulation models, including a multi-country macroeconomic simulation engine. (Most of this work is not in this GitHub account.)
• Quant interests — implied volatility surfaces, SVI / SABR fitting, HMM-based regime detection, portfolio optimization, Monte Carlo VaR.

These six repositories are chosen to show range across domains, not language stack.

The remaining repositories on this account are smaller experiments, learning artifacts, and tooling. See the full list at github.com/anonymousAAK?tab=repositories.

AI & Agentic Systems     ████████████████░░░░    aastf, agent harnesses, adapters
AI Security              ███████████████░░░░░    OWASP ASI 2026, prompt and tool injection
Quantitative Finance     ██████████░░░░░░░░░░    IV surfaces, SVI/SABR, HMM, Monte Carlo VaR
Scientific Computing     █████████░░░░░░░░░░░    Kepler data, MLP-CPU research
Languages & Compilers    ████████░░░░░░░░░░░░    adarshLang lexer/parser/interpreter
Systems & Dev Tooling    ████████░░░░░░░░░░░░    CLIs, sandboxes, CI integrations
Algorithms & Foundations ████████████░░░░░░░░    30-week curriculum, 4 languages

The bars represent relative emphasis in my public and semi-public work, not a skill self-rating.

I pick the language to match the problem, not the other way round.

For agents, ML, and analytics: Python, LangGraph, CrewAI, OpenAI Agents SDK, PydanticAI, FastAPI, Jinja, SARIF, OpenTelemetry.

For low-level and performance work: C++17, Rust.

For services and full-stack systems: Node.js, Express, React, TypeScript, MongoDB, MySQL, GraphQL.

For everyday correctness: Java, pytest, Hypothesis property-based tests, GitHub Actions, Linux, Bash.

I'm open to collaborations on hard, cross-disciplinary problems — AI safety and agent security, quantitative-finance research, scientific simulation, and developer tooling that takes correctness seriously. I'm interested in roles where systems thinking across domains is the job, not a side dish.

_{If something here resonates, send me a message. I prefer specific problems to generic intros.}