About Me
Independent researcher and software engineer with a math and physics background. I like understanding a system well enough to rebuild it, whether that's fluid dynamics, aircraft aerodynamics, or the AI tooling that helps me work through both faster.
cfd OpenFOAM, TENO/LES, turbulence closures
gui / viz Qt5, VTK, OpenCASCADE
systems Linux, WSL2/WSLg, CMake, GitHub Actions
About
I'm a recent 12th-grade graduate: a math and physics person who ended up writing a lot of code. Most of what I build starts from a first-principles question (how a fluid actually behaves, why an aircraft failed the way it did) and turns into software once I need to answer it at scale.
I'm based in Palakkad, Kerala, India, and split my time between independent research, software engineering, and CFD tooling. I'd rather derive something from scratch than take a black box's word for it.
All of it (the solver work, the fine-tuning, the analysis) was built on a single underpowered laptop. No workstation, no cluster. If the hardware's a bottleneck, that's a problem to work around, not a reason to wait.
Experience
Software Engineering Intern, Vittoba
Six months of hands-on software engineering experience in a production environment
Projects
CakeCFD
A full CFD workflow from CAD import through meshing, solving, and analysis. TENO convection schemes shipped as a wmake library, STEP/BREP geometry import via OpenCASCADE, a guided blockMesh/snappyHexMesh pipeline, and a native Qt5 + VTK viewport with live residual plots and streamline tracing. Also wires up direct Johns Hopkins Turbulence Database access for DNS inflow data. Two public repos, both GPLv3: the Cake Studio desktop app and the cakecfd_ai control layer.
cakecfd_ai
A Python package that exposes meshing, solving, monitoring, and post-processing as tools Claude can call: set up, run, and interpret a CakeCFD simulation from natural language, via a Python API, CLI REPL, or direct tool calls with no API key required.
Cross-Platform Installer
A packaging pipeline that ships the CFD suite natively on Linux and, on Windows, walks the user through WSL2 setup and consent, built and published automatically through a GitHub Actions → Releases pipeline.
jan-terminal-mcp
An MCP server that gives a local Jan AI model real hands on the machine: shell commands, git, file read/write, and code execution across 14 languages, entirely local and private. Ships with a 4B model fine-tuned via QLoRA specifically to call these tools reliably instead of just describing what it would do.
Skills & Tools
Research
Aerodynamic and Systems-Level Analysis of a 5th-Generation Fighter Using Classical Methods and Open-Source Data
USAF Digital DATCOM applied to open-source Su-57 geometry: stability derivatives, turn performance, LEVCON vortex-lift analysis
Inlet Aerodynamic Limitations and Thermal Failure Analysis of the HAL Tejas Mk1
A first-principles aerodynamic and thermodynamic investigation into the 2025 Dubai Airshow incident, built from published engine specs, NASA F404 exhaust survey data, and flight timeline reconstruction. Dedicated to the memory of Wing Commander Namansh Syal, Indian Air Force.
Mentioned in: idrw.org: "Researcher Flags Inlet Design Constraints in Tejas Mk1 After Dubai Incident"
Learning Experiences
Intermolecular Navier–Stokes Extension
A proposed thermodynamically-consistent extension to the Navier–Stokes equations for hypersonic continuum flow, adding an intermolecular pressure term via a van der Waals attractive contribution with Redlich–Kwong temperature scaling.
Submitted to the Journal of Fluid Mechanics and rejected after peer review.
What I learned from it
- The difference between a real scientific claim and an interesting one
- Scoping a research question against existing literature instead of in isolation
- Moderating claims to match what the evidence actually supports
- Which validation methods are accepted in the field, and which aren't
- What counts as an acceptable error margin for a claim like this
- How to properly structure a derivation for a reviewer, not just for myself
- A detailed review report, with the JFM deputy editor, Prof. Meneveau, personally investing time to guide the feedback
- The referees went beyond the rejection itself, giving detailed guidance for how to approach future submissions