Muhammed Shah

AI Software Engineer

About Me

I'm a software engineer who enjoys building applications and integrating AI/ML capabilities into them. I’ve started my career working in a startup environment and collaborated with experienced executives, which has given me a mix of hands-on development and exposure to high-level decision-making.


I’m always learning and currently diving deeper into neural networks, large language models (LLMs), and retrieval-augmented generation (RAG), while also starting to explore AI agents.

Work Experience View More →

Potentiam Ltd

Junior AI Software Engineer

Sept 2024 - Present

Working directly with the executive leadership building projects that leverage ML capabilities to meet specific functional requirements and ensuring smooth integration of LLMs into products that serve clients' needs.

AI Trainee Engineer

Sept 2023 - Sept 2024

Collaborating closely with the Global CIO and Founder/Director of our company, primarily focusing on AI-driven initiatives, while also being mentored to enhance my skills and expertise in the field.

Wipro Ltd

AI Intern

Mar 2023 - Jun 2023

Worked on developing and fine-tuning LLMs using LangChain, implementing a simple Q&A system over documents, optimizing its accuracy, and created an interface using Streamlit.

Work Projects View More →

Agentic AI Systems

  • Actively researching and prototyping Agentic AI architectures using tools like LangGraph, Google ADK, and Pydantic. The project explores the Google A2A protocol to enable autonomous, multi-agent interactions, currently in its early experimentation phase.

Private Enterprise GPT

  • Deployed and scaled an in-house Private GPT using an open-source model on an Ubuntu server.
  • Configured Nginx reverse proxy, SSL certification, and automated deployment with system services, cron jobs, and log rotation.
  • Migrated user authentication system to OAuth with Microsoft Authenticator, restricting access to employees under the company domain.
  • The solution is now used company-wide, replacing ChatGPT and DeepSeek while maintaining full control over data.

AI Search

  • Built an AI-powered search widget using a Retrieval-Augmented Generation (RAG) pipeline for internal tools at Potentiam Ltd.
  • Implemented FAISS for fast, dense vector indexing and used LangChain to connect document retrieval with LLM response generation.
  • Optimized embedding algorithms for better speed and relevance, and switched to document-based retrieval to reduce hallucinations.
  • Added caching layers to improve query response efficiency.
  • Achieved 90% semantic match accuracy and reduced user navigation time by 50%, confirmed through usage analytics and feedback.

Intelligence Framework

  • Designed and developed a modular AI framework to streamline Talent Acquisition and HR processes at Potentiam Ltd.
  • Delivered three main tools:
    • Resume Reviewer – Automated candidate evaluation using LLMs.
    • Job Spec Generator – Generated detailed job descriptions from simple prompts.
    • Resume Formatter – Standardized resume formatting using AI.
  • Enabled LLM-powered analysis and conversational UX for enhanced usability.
  • Reduced candidate screening time to under 10 seconds and data formatting/prep to 2–3 seconds, significantly boosting recruiter productivity.

Chatbot Development

  • Led development of an AI chatbot widget integrated into internal tools at Potentiam Ltd.
  • Migrated from Rasa Open Source to Rasa Pro CALM, enhancing stability and extending support for LLM-based fallback responses.
  • Integrated with Flask backend and overhauled the UI/UX for clarity and performance.
  • Delivered features like conversational flows, error fallback handling, and data exportation support.
  • Resulted in team-wide improvements, reducing internal task execution (navigation, info retrieval, data exports) to under a minute.

Skills

Programming Languages:

Python

ML & DL Frameworks:

PyTorch

HuggingFace

LangChain

LangGraph

AI Architectures:

Neural networks

LLMs

AI Techniques:

RAGs

NLU/NLP

Prompt engineering

AI Applications:

Chatbots (Conversational AI)

Private GPT

AI Agents

Vector Databases:

FAISS

Qdrant

Data Engineering & Analysis:

NumPy

Pandas

Matplotlib

Jupyter Notebook

AI Deployment & MLOps:

Linux Server (Ubuntu)

Ollama

Docker

AI Tools & Platforms:

Rasa

Open WebUI

Certifications View More →

External LinkBuilding Generative AI Applications Using Amazon Bedrock

Issued by Amazon Web Services (AWS) - Dec 2024

External LinkRasa Developer Certification

Issued by Rasa - Aug 2024

External LinkGenerative AI with Large Language Models

Issued by Deeplearning.ai - Feb 2024

External LinkData Science Nanodegree

Issued by Prepinsta - Dec 2023

External LinkPython Nanodegree

Issued by Prepinsta - Oct 2023

External LinkMachine Learning Nanodegree

Issued by Prepinsta - Aug 2023

External LinkCareer Essentials in Generative AI

Issued by Microsoft and LinkedIn - Jul 2023

Personal Projects View More →

External LinkRoad to AI

Developed this site as documentation of my learning journey in building Neural Networks and GPT models. It covers foundational concepts and advanced transformer-based architectures, serving as both a personal and shared knowledge base.

External LinkGPT Transformer Model-2

Building GPT-2 from Scratch: An end-to-end PyTorch implementation of a GPT-2 style language model (124M) released by OpenAI and inspired by Karpathy’s NanoGPT. Covers core components like tokenization, multi-head self-attention, transformer blocks, and positional embeddings. Implements key ML concepts such as causal masking, autoregressive modeling, and trains on real-world text for generation.

External LinkGPT Transformer Model-1

Building GPT from Scratch: A step-by-step implementation of a GPT-style language model, inspired by Andrej Karpathy’s lecture. The project covers language modeling fundamentals, self-attention mechanisms, and transformer architectures. It starts with a simple bigram model and builds up to a multi-headed self-attention transformer.

External LinkNeural Networks - Language Models

A hands-on exploration of character-level language models inspired by Karpathy's approach. This series covers the development of foundational ML models, the implementation of Multilayer Perceptrons based on seminal research papers, and experimentation with advanced techniques such as backpropagation, language modeling, and neural network concepts.

External LinkNeural Networks - Micrograd

Built a neural network from scratch by developing a micrograd library, implementing core concepts like backpropagation and gradient descent following Andrej Karpathy’s methodology.

External LinkPortfolio website with Chatbot

Developed a Portfolio site using Flask, HTML, CSS and JavaScript. Along with a functioning chatbot developed using RASA.

Education View More →

MVJ College of Engineering (Affiliated with VTU), 2023 Graduate

BTech - Computer Science

Grade: 9.32 CGPA

Achievements: Top 10 University Rank Holder in CSE Department, Director of Design at TEDxMVJCE, Vice President at Saahitya Literature Club

Primus PU College, 2019 Graduate

PUC - Physics, Chemistry, Math and Computer Science

Grade: 86.5%, Distinction

St.Peter's School, 2017 Graduate

I.C.S.E - Science

Grade: 92.4%

Achievements: Top Rank Holder in Computer Applications.