AI Software Engineer
Hello! I’m Muhammed Muzzammil Shah. I design and engineer intelligent systems by leveraging LLMs and integrating AI architectures. I like implementing practical AI solutions, whether it’s developing AI-powered chatbots, agentic systems, or adding intelligent capabilities to existing software.
My expertise includes AI techniques and architectures like Agentic AI, RAG and prompt engineering, and I enjoy working with frameworks such as PyTorch, Hugging Face, and LangChain. I also have a good foundation in full-stack development, with skills in Python, React.js, and REST APIs.
I’m dedicated to expanding my knowledge and sharing what I learn through my Road to AI documentation. Feel free to explore it as I share what I learn about AI in clear, practical terms.
Potentiam Ltd
Sept 2024 - Present
Building and optimizing applications using LLMs or SMLs for different departments and also integrating AI architectures for secure, scalable deployment. Currently working on upgrading these systems to include agentic components.
Sept 2023 - Sept 2024
Collaborated directly with the Global CIO and Founder on AI-driven initiatives, focusing on open source AI R&D, pilot implementations and cybersecurity operations.
Wipro Ltd
Mar 2023 - Jun 2023
Developed a pilot for interactive data insight conversations and researched AI observability, ML monitoring, root cause analysis, anomaly detection, and capacity forecasting.
Designed, configured, and deployed a secure in-house Private GPT system on Ubuntu with Nginx reverse proxy, SSL, and custom domain mapping. Integrated Microsoft OAuth authentication to ensure access only to company users, automated backend services for resilience, and enabled GPU powered remote access via VNC. Migrated all AI interactions in-house, ensuring full data privacy and operational reliability.
Developed an AI powered search widget with an “Ask AI” feature for instant, accurate querying of company salary and HR data, leveraging a RAG system with FAISS embeddings, LangChain orchestration, and Llama3.x models. Optimized embedding accuracy, reduced hallucinations, and introduced caching for speed, deploying the solution as a JavaScript widget via IIS for seamless internal integration.
Engineered an all-in-one intelligence framework for HR and Talent Acquisition, featuring a LLM driven resume reviewer, job spec generator, chatbot, and resume formatter with advanced UI and batch processing. Utilized local LLMs via Ollama, LangChain pipelines, and MongoDB for fast, secure, and accurate candidate evaluation, automation, and internal data generation.
Led the creation of an advanced company chatbot widget, powered by a proprietary SLM trained on internal data and later enhanced with LLM (Ollama) fallbacks. Delivered features like file downloads, speech-to-text, custom UI, and robust conversational management. Migrated from Rasa Open Source to Rasa Pro CALM for advanced context switching and ensured scalable, low latency responses for dynamic business needs.
Building Generative AI Applications Using Amazon BedrockIssued by Amazon Web Services (AWS) - Dec 2024
Rasa Developer CertificationIssued by Rasa - Aug 2024
Developing AI Applications with Python and FlaskIssued by IBM - Mar 2024
Generative AI with Large Language ModelsIssued by Deeplearning.ai - Feb 2024
Data Science NanodegreeIssued by Prepinsta - Dec 2023
Python NanodegreeIssued by Prepinsta - Oct 2023
Machine Learning NanodegreeIssued by Prepinsta - Aug 2023
Career Essentials in Generative AIIssued by Microsoft and LinkedIn - Jul 2023
Developed a comprehensive documentation site covering practical neural network and GPT model implementations, foundational ML concepts, and advanced topics like transformers and tokenization, using Python and PyTorch.
Built and deployed a React-based AI assistant on my personal website using the AI SDK, enabling multi-step reasoning, tool use, and real time responses with seamless UI/UX and dynamic knowledge updates.
Implemented a full GPT-2 style language model in PyTorch from first principles, covering all key components such as tokenization, self-attention, transformer blocks and trained it on real world text for natural language generation.
Created a GPT style transformer in PyTorch from scratch, progressing from a simple bigram model to a full transformer with multi-head attention, feedforward layers, and text generation capabilities.
Built neural networks of different language models progressing from a basic bigram and MLP-based character level predictor to deeper architectures with batch normalization, manual backpropagation, and a WaveNet inspired convolutional network.
Built a neural network library from the ground up in Python, implementing backpropagation, gradient descent, and autograd features to understand neural network fundamentals.
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
PUC - Physics, Chemistry, Math and Computer Science
Grade: 86.5%, Distinction
I.C.S.E - Science
Grade: 92.4%
Achievements: Top Rank Holder in Computer Applications.