Selected work · 2024–2026
Different projects. One consistent instinct: understand the system.
From an offline ASL translator and award-winning education platform to malware triage, source-linked RAG, and a 64-bit operating system. Each case study explains the decisions—not just the stack.
Flagship work
The four projects that best represent my systems depth, AI engineering judgment, machine-learning foundation, and product execution.
boot: entering long mode
gdt: loaded
paging: initialized
kernel: ready
akkash-os> _Systems programming · OS internals
64-bit operating system
A small operating system that boots from hand-written assembly into a 64-bit C kernel and interactive shell.
- C
- x86 Assembly
- QEMU
- GDT
STATIC ANALYSIS
01 File type + PE headers verified02 YARA indicators verified03 Model interpretation inferredNO EXECUTION · FAIL CLOSEDAI product · Security tooling
Malware triage tool
A safety-first analysis pipeline that turns static evidence and external lookups into a structured, plain-English report.
- Python
- React
- Claude API
- YARA
Machine learning · Computer vision
Radiography classification
A multi-class chest X-ray classifier paired with an interpretability dashboard for non-technical review.
- TensorFlow
- Keras
- MobileNetV2
- Streamlit
STEM FORGE
What do you want to learn?
AI education · Full-stack product
STEM Forge
An award-winning, local-first learning platform that generates adaptive quizzes, coding problems, and custom courses.
- Python
- FastAPI
- React
- Gemma 3n
More builds
Local AI, agent evaluation, accessibility, retrieval, developer tools, Linux, networking, and human-centered product work.
CODIVUS
Programming help, running locally.
AI education · Offline-first software
Codivus
An offline AI programming-education platform combining local language models with retrieval-augmented learning workflows.
- React
- FastAPI
- Ollama
- RAG
AI agents · Evaluation
BenchMind
An agent-benchmarking platform built during Cal Hacks to make AI-agent behavior easier to compare and reason about.
- AI Agents
- Benchmarking
- Evaluation
- Hackathon
AI systems · Performance engineering
Open-weight LLM benchmarking
A reproducible evaluation framework for choosing local language models based on capability and inference trade-offs.
- Python
- Ollama
- Gemma
- DeepSeek
Multimodal AI · Accessibility
ASL translator
An offline-first multimodal pipeline translating webcam ASL gestures into English text and speech.
- OpenCV
- Gemma 3n
- FastAPI
- Ollama
Retrieval systems · Data engineering
YouTube RAG
A transcript-ingestion and retrieval service where every generated answer points back to its source and timestamp.
- Python
- FastAPI
- OpenAI
- Milvus
WRITING PLAN
Interpret the prompt Build an argument Support with evidence Revise for clarity68% completeAI product · Developer tooling
Scaffold
A collaborative writing assistant that decomposes rubrics into tasks and gives feedback as a document evolves.
- TypeScript
- Claude API
- Browser Extension
- Google Docs
CURAPOINT
01 Understand the need02 Find the next step03 Move forward clearlyHEALTHCARE ACCESSHealthcare access · Product engineering
CuraPoint
A healthcare-access application exploring how software can make a complicated, high-friction journey easier to navigate.
- Product Design
- Healthcare Access
- Application Development
systemctl: active
journal: clean
hyprland: ready
permissions: verified
aakkash@arch ~ $ _Linux · Systems operations
Arch Linux environment
A hand-configured Arch Linux daily driver with Hyprland, custom services, logging, permissions, and automated setup.
- Arch Linux
- Hyprland
- systemd
- Bash
ENROLLMENT NETWORK
Software engineering · Networking
Enrollment system
A Java client-server application supporting concurrent students, instructors, and administrators with role-based permissions.
- Java
- TCP/IP
- Sockets
- Multi-user architecture