Email • LinkedIn • GitHub

Welcome to my corner of the AI world! I’m Jawad, an AI Engineer and CS student who loves taking ideas from paper → prototype → real, usable tools.

I see programming as a way to build intelligent systems—LLMs, RAG pipelines, and agents—that actually help people, not just live in notebooks.

• AI Engineer & Founder of AI by MJR
  Building LLM systems, RAG pipelines, and agentic workflows that go from concept to production.

• CS @ University of Central Punjab
  CGPA 3.77 / 4.0 with strong foundations in algorithms, systems, and machine learning.

• Industry Experience
  Currently working as an AI/ML Intern (DeveloperHub Corporation®, CodeAlpha) on:

  • LLM automation
  • Chatbots & conversational AI
  • Computer vision & applied ML

• Competitive Programming & Problem Solving
  Winner – Riphah RC3 Speed Programming Competition.

• Self-Learner
  Constantly exploring modern Transformer architectures, agentic AI patterns, and better ways to evaluate & ship GenAI systems.

• 🧠 A GPT-like LLM from scratch in PyTorch
  Tokenizer, attention, sampling, training loop, and training utilities.
  Progress:

• 💬 RAG chatbots (LangChain + ChromaDB + local Phi via Ollama)
  Document Q&A over custom knowledge bases, with clean retrieval pipelines.
  Progress:

• 🤖 Agentic workflows with LangGraph + AgentOps
  Multi-step tool use, routing, and observability for production-grade agents.
  Progress:

• 🎓 Educational repos & examples
  Simple, well-documented projects to help students get into GenAI & LLM engineering.
  Progress:

Languages
Python · C++ · SQL

ML / DL
PyTorch · NumPy · Transformers · Attention Mechanisms · CNNs · RNNs · Training Loops · Autograd

LLM / GenAI
LangChain · LangGraph · FAISS · ChromaDB · OpenAI API · Ollama · RAG · Agents · Evaluation

Backend & Apps
FastAPI · Streamlit · Flutter

Tools & Infra
Docker · Git · Linux

Here are some projects that represent the kind of AI work I like doing:

• LLM-From-Scratch
  GPT-style Transformer implemented from scratch in PyTorch.

  • Custom tokenizer, positional encodings, multi-head attention
  • Autoregressive decoding and sampling (top‑k, temperature)

• RAG Chatbot with LangChain + Ollama Phi
  Retrieval-augmented chatbot over custom knowledge bases.

  • ChromaDB for dense retrieval & embeddings-based search
  • Local inference with Phi via Ollama + Streamlit UI

• GenApp Tool Agent & Trending Agent
  Agentic workflows to generate and refine app ideas.

  • Built with LangGraph (routing, memory, tool-calling)
  • Observability and debugging with AgentOps

• Micrograd-style Autograd Engine
  Minimal deep-learning core built from scratch in Python.

  • Computational graph & backprop implementation
  • Used to train small MLPs and understand DL internals

I keep improving these, so check back for updates, new branches, and more agents / RAG examples.

• Scaling RAG and agentic systems (retrieval quality, orchestration, evaluation)
• Optimizing LLM inference: quantization, caching, batching
• Better MLOps for GenAI: observability (LangSmith / AgentOps), experiment tracking, reproducible pipelines
• Connecting LLM backends with clean frontends (Streamlit, Flutter, etc.)

• Going deeper into:
  • Transformer internals & training stability
  • Retrieval strategies for long-context and enterprise RAG
  • Patterns for robust, tool-using agents

I like turning what I learn into small, focused repos—so others can follow the same path into AI engineering.

• I enjoy turning complex AI concepts into simple, practical code that others can learn from.
• I love hackathons, fast prototyping, and building in public under AI by MJR.
• I see every bug as a hint that there’s something new to understand.

If you’re working on:

• LLM systems (RAG, agents, evaluation, optimization)
• GenAI products & tooling
• Applied ML in education, productivity, or developer tools

I’d love to connect or collaborate.