Skip to content
View Parv-01's full-sized avatar
⏳
Focusing
⏳
Focusing

Highlights

  • Pro

Block or report Parv-01

Report abuse

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

Report abuse
Parv-01/README.md
Parv Agarwal
Personal Website   Profile Views   Followers   Coffee Chat

Typing SVG

Attack on Titan β€” Parv Agarwal opening banner

divider

βš”οΈ   CHARACTER PROFILE   βš”οΈ

γ€Œ The one who knows the most wins. The one who ships it, survives. 」


# ──────────────────────────────────────────────────────────────────────
#  PARV-01  //  PERSONNEL FILE  //  CLEARANCE: UNRESTRICTED
# ──────────────────────────────────────────────────────────────────────

Character:     "Parv Agarwal"
Code:          "PARV-01"
Website:       "https://parvagarwal.is-a.dev"

Affiliations:
  Research:    "Junior Research Associate β€” AI-NLP-ML Lab, IIT Patna"
  Teaching:    "CS Faculty β€” Poornima University (DSA, DL, Data Science, Cloud)"
  Industry:    "Quantum Course Author β€” ChipXpert Technologies"
  Online:      "Educator Β· Content Creator Β· Freelancer"

Class:         "AI/ML Researcher Β· Full-Stack Dev Β· Quantum Engineer Β· Educator"
Level:         "Mid β†’ Senior (closing the gap at speed)"

Status:        "πŸ”΄ Building on three research frontiers simultaneously"
Research_Arc:  "Low-Resource NLP Γ— Neuromorphic LLMs Γ— Quantum-Classical Hybrids"

Research_Focus:
  Primary:
    - "Low-Resource NLP β€” making language models work for everyone, everywhere"
    - "Neuromorphic LLMs β€” brain-inspired, spike-based approaches to language"
    - "Quantum-Classical Hybrid architectures β€” where physics meets ML"
  Secondary:
    - "Embedded Electronics and Edge AI on constrained hardware"
    - "Brain-Inspired Computing for efficient inference"

Engineering_Arsenal:
  - "LLM Fine-tuning via LoRA, QLoRA, RLHF and DPO β€” end to end"
  - "RAG pipelines with LangChain, LlamaIndex, vector DBs"
  - "Full-stack web apps from React to FastAPI to production deployment"
  - "Quantum Circuit Design with Qiskit and local Aer Simulation"
  - "Tokenisation from scratch β€” BPE, WordPiece, SentencePiece, Tiktoken"

Currently_Pushing:
  - "Quantum-Classical Hybrid Model architectures that actually run locally"
  - "Neuromorphic approaches to transformer efficiency"
  - "Agentic LLM systems with real-world tool use"
  - "Growing an audience that trusts what I build and teach"

Available_For:   "Research collabs Β· Freelance builds Β· Teaching Β· Coffee chats"
Known_Weakness:  "Rabbit holes with no exit Β· Bugs at 3AM Β· Too many tabs open"
Battle_Cry:      "I will keep moving forward."

divider

🧬   RESEARCH SPOTLIGHT   🧬

γ€Œ Research at the intersection of classical ML, quantum computing and neuromorphic systems. 」
Read more at parvagarwal.is-a.dev


Area What it is Why it matters
🧠 Low-Resource NLP Building language models that work without massive compute, data, or budget Most of the world's languages don't have billion-token corpora. Most developers don't have A100s. This research asks how NLP can work for everyone, not just the well-resourced.
⚑ Neuromorphic LLMs Brain-inspired, spike-based computing architectures applied to language models The human brain runs on roughly 20 watts. GPT-4 training used gigawatt-hours. Neuromorphic systems process information event-driven, asynchronously β€” potentially cutting the energy cost of inference by orders of magnitude.
βš›οΈ Quantum-Classical Hybrids Circuits that combine quantum operations with classical ML layers Full quantum computers are years away from practical NLP use. Hybrid architectures β€” quantum feature maps, variational circuits, quantum attention β€” give us a way to explore quantum advantage now, on local simulators, with real algorithms.
πŸ”Œ Embedded Electronics + Edge AI Running intelligence on microcontrollers, FPGAs, and constrained systems A model that only runs in the cloud is a model that fails when connectivity does. Edge AI means the intelligence travels with the device. This intersects directly with low-resource NLP and neuromorphic efficiency work.


divider

πŸ”₯   THE ARSENAL   πŸ”₯

γ€Œ A soldier goes to war with the weapons they have mastered, not the ones they wish they had. 」


L from Death Note β€” deep in thought
γ€Œ I am the world's greatest detective. I know exactly which model to fine-tune and which vector DB to reach for. 」

Languages

Languages



AI Β· Machine Learning Β· Deep Learning

ML Frameworks    HuggingFace   LangChain   LlamaIndex   Unsloth



LLM Infrastructure Β· APIs Β· Serving

OpenAI   Anthropic Claude   Ollama   vLLM   Groq   Mistral   Together AI



Vector Databases Β· RAG Β· Embeddings

ChromaDB   Pinecone   Qdrant   Weaviate   FAISS



Quantum Computing Β· Neuromorphic

Qiskit   IBM Quantum   AerSimulator   PennyLane   Brian2 Spiking Neural Network   snnTorch   FakeNairobiV2



Embedded Β· Edge Β· IoT

Embedded    ONNX Runtime   TensorFlow Lite   Edge Impulse



Web Β· Full Stack

Web Stack



Databases Β· Storage

Databases



Infrastructure Β· DevOps Β· Cloud

DevOps



Research Β· Experiment Tracking Β· Productivity

Dev Tools    W&B   MLflow   LaTeX   Obsidian

divider

πŸ›    WHAT I BUILD AND HOW I CAN HELP   πŸ› 

γ€Œ Every wall is just a problem that hasn't been solved yet. 」


If you need this I can actually do this
🧠 AI / LLM System End-to-end RAG pipelines, LLM fine-tuning (LoRA/QLoRA/DPO), agent architectures, custom chatbots with memory and multi-step tool use
βš›οΈ Quantum Computing Qiskit circuit design, quantum algorithm walkthroughs, hybrid classical-quantum implementations β€” all running locally, no IBM hardware bill
🧬 Research Collaboration Low-resource NLP experiments, neuromorphic model prototyping, quantum-classical hybrid architectures β€” IIT Patna research context
πŸ”Œ Edge AI / Embedded Running ML inference on Raspberry Pi, Arduino, FPGAs β€” ONNX, TFLite, quantized models, model compression
🌐 Full-Stack Web App React/Next.js + FastAPI or Node backend, PostgreSQL schema, deployment on cloud or VPS β€” design to production
πŸ“š Technical Writing / Education Course outlines, lecture scripts, deep-dive documentation, explainers for complex topics that actually make sense
πŸ’¬ Technical Consulting Architecture reviews, model selection, stack decisions β€” direct, honest, opinionated. No fluff.
✍️ Freelance Builds ML APIs, data pipelines, internal AI tools, developer docs, proof-of-concept builds with real deliverables

divider

πŸ“Š   BATTLEFIELD STATS   πŸ“Š

γ€Œ Numbers don't lie. The battlefield never does. 」


Parv's GitHub Stats    Top Languages

GitHub Streak

GitHub Trophies

Activity Graph

Contribution Snake Animation

divider

🎯   ACTIVE MISSIONS   🎯

γ€Œ A soldier who doesn't advance is already dead. Forward. Always forward. 」


Survey Corps β€” charging into the unknown
γ€Œ Dedicate your heart. Deploy your code. Publish your findings. 」



Mission Type Status Stack
πŸ”¬ Quantum Computing Course β€” Module 3: Qiskit and Quantum Programming, ChipXpert Technologies Education πŸ”΄ Active Qiskit Β· Aer Β· Jupyter
🧬 Low-Resource NLP Research β€” experiments in constrained-compute language modelling at IIT Patna AI-NLP-ML Lab Research πŸ”΄ Active PyTorch Β· HuggingFace Β· W&B
πŸ€– Agentic LLM System β€” multi-step tool-use agent with persistent memory and structured retrieval Engineering 🟑 Building LangChain Β· LlamaIndex Β· Ollama
🧠 Neuromorphic LLM Exploration β€” spike-based approaches to attention and language representation Research 🟑 Exploring snnTorch Β· Brian2 Β· PyTorch
🌐 Personal Portfolio β€” parvagarwal.is-a.dev Web Dev 🟒 Live HTML Β· CSS Β· JS
πŸ“ ML Mathematics Notes β€” graduate-level synthesis covering linear algebra, calculus, probability Education 🟒 Complete Python Β· ReportLab Β· LaTeX
🧩 RAG Pipeline Architecture Guide β€” production reference for LangChain retrieval and vector search Docs 🟒 Complete LangChain Β· FAISS Β· Ollama
πŸ”§ Tokenisation From Scratch β€” BPE, WordPiece, Unigram, SentencePiece, Tiktoken all implemented Research + Code 🟒 Complete Python Β· HuggingFace
🎀 Quantum Lecture Series β€” 4-lecture series from gate mathematics through Qiskit fundamentals Education 🟒 Complete Markdown Β· PowerPoint

divider

πŸ““   FROM THE PARV-01 NOTE   πŸ““

γ€Œ The human whose code I review shall not be roasted publicly β€” unless it truly deserves it. 」


Death Note

╔══════════════════════════════════════════════════════════════════════╗
β•‘                                                                      β•‘
β•‘                    ✦   THE  PARV-01  NOTE   ✦                       β•‘
β•‘                                                                      β•‘
╠══════════════════════════════════════════════════════════════════════╣
β•‘                                                                      β•‘
β•‘  Rule I      The code I ship shall be deployed. There are no        β•‘
β•‘              exceptions. If it never shipped, it never existed.     β•‘
β•‘                                                                      β•‘
β•‘  Rule II     This note shall not take effect unless the writer      β•‘
β•‘              has actually run the tests at least once in their      β•‘
β•‘              life before opening a pull request.                    β•‘
β•‘                                                                      β•‘
β•‘  Rule III    Every bug has a name and a face. I will find it.       β•‘
β•‘              I always find it. It is only a matter of time.         β•‘
β•‘                                                                      β•‘
β•‘  Rule IV     The commit message must be meaningful. Writing          β•‘
β•‘              "fix stuff" is not a commit message. It is a           β•‘
β•‘              distress signal. I am sending help.                    β•‘
β•‘                                                                      β•‘
β•‘  Rule V      If you open a pull request without documentation,      β•‘
β•‘              you have already lost. The docs are the product.       β•‘
β•‘                                                                      β•‘
β•‘  Rule VI     I'll take a dataset and analyze it.                    β•‘
β•‘              The truth was in the distribution all along.           β•‘
β•‘                                                                      β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

divider

β˜•   GRAB A COFFEE WITH ME   β˜•

γ€Œ The best conversations happen over coffee. The best research ideas happen right after. 」


I genuinely enjoy conversations that go somewhere β€” whether that's your next AI project, a research rabbit hole you've fallen into, a freelance problem you need a second opinion on, or just what you think is going to happen with quantum computing and neuromorphic systems in the next five years. If any of the following sounds like you, send me a message. I respond to everyone.


You're thinking about... Let's talk
πŸ€” Building an AI product but unsure where to start I've scoped and shipped these. Let's map it out.
πŸ”¬ A research problem in NLP, quantum, or neuromorphic computing I work on these at IIT Patna. Let's dig in.
πŸ“– Learning LLMs, RAG, Qiskit, or neuromorphic ML I teach this. Ask me literally anything.
πŸ’» Need a full-stack app with AI built in I build these. Let's scope it together.
πŸ“ Want your technical writing to actually land I write and edit these. Send it over.
πŸŽ“ Confused about the AI/ML/Quantum path to take I've been there. Happy to share what I know.
🌐 Want to see what I'm currently working on Visit parvagarwal.is-a.dev β€” it's all there.
β˜• Just want to talk tech, research, life, or all three Best reason of all.

divider

🌐   FIND ME ACROSS THE WALLS   🌐

γ€Œ The walls won't stop information from moving. Neither will I. 」


Portfolio   LinkedIn   Medium   Email   Twitter/X



"The only walls that matter are the ones you build inside your own head."

β€” Parv Agarwal, probably at 2AM debugging a quantum circuit with cold coffee, three terminals open, and a neuromorphic paper on the side

footer β€” Dedicate Your Heart

Pinned Loading

  1. gpualert gpualert Public

    Monitor GPU jobs, detect failures, and stay informed during long-running ML experiments.

    Python 4

  2. parv-portfolio parv-portfolio Public

    This is repo to keep on evolving my personal web portfolio with intent of keeping it always free of charge.

    TypeScript

  3. NYUAD-2022 NYUAD-2022 Public archive

    Forked from affifboudaoud/NYUAD-2022

    Jupyter Notebook

  4. LLM-Hugging-face-Basics LLM-Hugging-face-Basics Public

    HTML

  5. IItp-Celesta/celesta-25-site IItp-Celesta/celesta-25-site Public

    JavaScript 2 10

  6. iitpanwesha/anwesha-frontend iitpanwesha/anwesha-frontend Public

    Forked from DEDSWIN/frontend

    Frontend of Anwesha 2026

    JavaScript 11