Skip to content
View aksaN000's full-sized avatar
🕵️
Learning
🕵️
Learning
11 followers · 5 following

Achievements

Achievement: YOLOAchievement: Pull Shark

Achievements

Achievement: YOLOAchievement: Pull Shark

Highlights

  • Pro

Block or report aksaN000

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
aksaN000/README.md

Md. Aksan Gony Alif: AI research, systems and web. Final-year CSE, BRAC University, Dhaka.

I. PrefaceHello, I'm Aksan, someone who has to know how a thing works before I'm comfortable using it. When I build something like a multithreaded process manager in C, I end up following its mutexes and spinlocks down to the assembly. My thesis started the same way, from my own intuition about memory and confidence in language models. I like reasoning from first principles. More than once I've worked my way to an idea only to find it already in the literature, which taught me to read earlier and build on what's there. Where I'm headed: a research-capable, systems-aware, full-stack AI engineer.

II. Current workThesis: CAEM, Confidence-Aware Episodic Memory, reducing hallucination in LLMs. Building: Root to End, interactive simulations tracing systems concepts from the kernel up to the code. Writing: From Zero, a book on the layer of computing a CS degree assumes but rarely teaches. Next: an image-processing project. Exploring: AI engineering, research and web development. Planned: a hobby operating system in C, from scratch.

III. Selected works

CAEM: confidence-aware episodic memory that reduces LLM hallucination SAM 2 for Sentinel-1 SAR flood mapping Root to End: interactive simulations tracing systems concepts from the kernel up to the code Continual learning: EWC vs. experience replay for BERT Text classifiers compared: classical ML, RNNs and transformers across 25 experiments Process Manager: simulated OS process table in C with POSIX threads Seven VAEs for multilingual music clustering MemeStack: full-stack meme studio and community

IV. InstrumentsLanguages: Python, C, JavaScript, TypeScript, PHP, LaTeX, Bash. Machine learning: PyTorch, Hugging Face, Lightning, scikit-learn, pandas, NumPy, W&B, Jupyter. Web: React, Node.js, Express, MongoDB, MySQL, MUI, Bootstrap. Tooling: Git, Linux, Vercel.

V. MarginsPlaying: Valorant, PUBG, CS2 and Clash Royale. Writing: personal journals, in LaTeX. Someday: a long list of places to see.

VI. Correspondence

Email: aksangoni.alif@gmail.com LinkedIn GitHub: @aksaN000 Facebook

Set in Cormorant Garamond and JetBrains Mono. Composed in Dhaka.

Pinned Loading

  1. Root-to-End-of-Random-Technical-Concepts Root-to-End-of-Random-Technical-Concepts Public

    Interactive simulations that trace CS concepts from the root to the end: the JS event loop, V8, an LR parser, fetch() to the wire, HTML to pixels, a click to its listener, threads vs. epoll. Built …

    HTML

  2. Continual-Learning Continual-Learning Public

    EWC vs. experience replay for catastrophic forgetting in a BERT intent classifier trained sequentially on HWU64 domains; 16-config sweep with 12 metrics. Stack: Python, PyTorch, HF Transformers.

    Python

  3. car-destroyer-game car-destroyer-game Public

    Car Destroyer Game: a two-way-traffic arcade shooter with boss fights, local two-player mode and particle explosions, drawn entirely with midpoint line and circle algorithms. Stack: Python, PyOpenG…

    Python

  4. A-Comparative-Analysis-of-Traditional-Machine-Learning A-Comparative-Analysis-of-Traditional-Machine-Learning Public

    25 experiments comparing classical ML, RNNs and pre-trained transformers on 10-class question/answer text classification. Stack: Python, scikit-learn, PyTorch, Hugging Face Transformers.

    Jupyter Notebook

  5. caem-thesis caem-thesis Public

    CSE400 thesis (BRAC University): confidence-aware episodic memory + self-improvement that cuts LLM hallucination (+20.3% EM, -29.4% hallucination vs. baseline). Stack: Python, PyTorch, Qwen-2.5-3B,…

    Python

  6. process-manager process-manager Public

    Simulated OS process table in C: concurrent fork/exit/wait/kill across POSIX threads, zombie reaping, orphan adoption by init, and an event-driven snapshot monitor. Stack: C, pthreads, ThreadSaniti…

    C