Mohammed Abul Fazal Fazlullah Sharief

Summary

Backend & AI Engineer with 3+ years building reliable, scalable backend and AI systems. Currently at Crowe, designing the platforms and tooling that let AI & ML Engineering teams ship reliable AI products at scale. Background in distributed architectures, high-throughput pipelines, and cloud-native systems on AWS across PostgreSQL, MongoDB, DynamoDB, and Redis. Deep hands-on experience with LLMs, RAG, AI Agents, and Vector Search — from architecture to production deployment.

Skills

• Programming Languages: JavaScript, TypeScript, Python
• Frameworks/Runtimes: NodeJS, FastAPI, Django
• Databases: PostgreSQL, MongoDB, DynamoDB, Redis
• Cloud Technologies: Amazon Web Services - EC2, ECS, S3, SQS, DynamoDB, Bedrock, Lambda, Aurora PostgreSQL Serverless, IAM
• AI/ML: LangChain, LangGraph, Large Language Models(LLM), Retrieval Augmented Generation(RAG), AI Agents, Vector Embeddings, Vector Search, scikit-learn, TensorFlow
• Tools: Git, Docker, Terraform, Kafka, GitHub Actions
• Methodologies: Agile, Scrum, Object-Oriented Programming (OOP)
• Testing: Jest, PyTest

Experience

• AI Platform Engineer | Crowe | May, 2026 - Present
  • Building AI Studio.
• SDE - 2, Backend & AI | Early Stage Startup(Stealth) | June, 2025 - January, 2026
  • Designed and built Backend Services, AI Agents, and RAGs.
• Software Engineer, Backend & AI | DAZN | June, 2024 - June, 2025
  • Designed and built independently DAZN’s first Large Language Model (LLM) application, leveraging generative AI with Retrieval Augmented Generation (RAG) using Vector Search, that provides real time analysis of customer support queries, helping the operations team derive strategies to resolve them quickly.
  • Implemented Parallel Computing in NodeJS to split thousands (140K+) of high computing tasks, reducing overall processing time from 24 hours to less than 5 hours.
  • Implemented Byte Order Mark (BOM) for CSV files to persist UTF-8 encoding in applications like MS-Excel.
  • Set up SQS for consuming out-of-order messages via a centralized streaming service based on SNS, storing results in DynamoDB.
  • Deployed Meta’s Llama 3.2 on AWS Bedrock and built an AI writing assistant for agents.
  • Integrated HMAC (Hash-based Message Authentication Code) APIs.
• Junior Software Engineer, Backend & iOS | Vigocare | Dec, 2022 - June, 2024
  • Reduced a MongoDB collection query time from 45sec to 3sec using projection, saving thousands of dollars in caching costs.
  • Developed and automated publishing, installation and deletion of a private npm package, resulting in a 10x reduction in code repetition and bugs.
  • Integrated a third-party video calling service and developed all APIs from scratch.
  • Migrated the mongoose library from v5 to v8 (3 major versions) across services.
  • Improved an API’s performance by 40x, reducing response size from 20MB to 500KB, resolving a gRPC bottleneck.
  • Contributed to the iOS app occasionally, integrating Core Data and multiple new APIs.

Projects

• Attention | GitHub Repo

  • An analytical tool leveraging a pre-trained BERT language model to predict masked words and generate self-attention visualizations. Makes transformer internals interpretable by mapping how attention heads distribute focus across tokens.

• Traffic Sign Recognition | GitHub Repo

  • A Computer Vision solution achieving 98% accuracy using Convolutional Neural Networks (CNN). Developed a multi-layer architecture for real-time classification of 43 categories of traffic signs based on the German Traffic Sign Recognition Benchmark (GTSRB).

• Orderbook | GitHub Repo

  • A high-performance multi-symbol orderbook service engineered with Node.js, TypeScript, and PostgreSQL. Features an automatic trade matching engine for limit and market orders, processing and settling transactions in real-time.

• Nim AI | GitHub Repo

  • A reinforcement learning agent that masters the game of Nim via Q-learning. The AI trains by playing thousands of games against itself, iteratively improving its strategy through standard Q-value updates and exploring the mathematical properties of the game.

• RAG Microservice | GitHub Repo

  • A specialized Retrieval-Augmented Generation service built with FastAPI and LangChain. Integrates FAISS vector search with security layers to prevent prompt injections while providing context-aware answers from custom documents.

• Yoga Pose Recommender | GitHub Repo

  • An AI-powered semantic search system using Gemini for context-aware descriptions and vector embeddings. Built with Node.js and Firestore, it enables high-performance natural language queries and accessibility features via Google Cloud Text-to-Speech.

• View all Projects.

Education

• Bachelor of Engineering, Electrical & Electronics | Osmania University | July, 2021

Blogs

• What happens when you open an app | September, 2024
  • Ever heard that mysterious ‘zzzzzz’ from your CPU when you launch a big app? Here’s what’s going on behind it.
• What happens when you click this link | June, 2024
  • Everything you need to know about what goes under the hood when you click a link or type a URL in the browser.
• How to take Harvard University’s CS50 Course and break into the Tech Industry | March, 2024
  • In the 21st Century, if there is anything that has comparatively added more value to the world than any other thing, it is software. The reason you’re reading this from any corner of the world, regardless of where it was published, is all because of the software.

Coursework

• CS50X: Introduction To Computer Science | Harvard University | Certificate
• CS50W - Web Programming with Python and JavaScript | Harvard University | Certificate
• CS50AI - Introduction to Artificial Intelligence with Python | Harvard University | Certificate
• Fundamentals of Network Engineering - Hussein Nasser | Udemy | Certificate