Sheroze Ajmal

Sheroze Ajmal

Backend Developer & ML Engineer

About Me

A passionate developer with over a decade of coding experience, starting from age 12. Currently working at BudSense, focusing on AI integration and internal tooling development. Previously founded and successfully exited a tech startup.

Quick Facts

  • 🚀 Built and sold a successful tech startup
  • 💡 Specialized in AI/ML and distributed systems
  • 🌐 Worked with clients across 4 continents
  • 📚 Continuous learner and tech enthusiast

Current Focus

  • Advanced RAG systems optimization
  • Large-scale distributed systems
  • AI-powered developer tools

Beyond Code

  • 🎥 Cinematography and color grading enthusiast
  • 🎸 Amateur audio programming
  • 🏃‍♂️ Fitness and outdoor activities

Technical Skills

Frontend Development

  • React & Next.js
  • Vercel Deployment
  • Modern UI/UX Design
  • Performance Optimization

Backend Development

  • Node.js & Express
  • API Development
  • System Design
  • Database Architecture

Machine Learning

  • Large Language Models
  • PyTorch & TensorFlow
  • ML Operations
  • AI Integration

Cloud & DevOps

  • AWS & Cloudflare
  • Docker & Kubernetes
  • CI/CD Pipelines
  • Infrastructure as Code

Featured Projects

Koko - AI-Powered Personal Growth Platform

Built a sophisticated AI companion platform demonstrating expertise in:

AI & ML Implementation

  • Custom LLM fine-tuning for personalized interactions
  • Emotional intelligence modeling using advanced NLP
  • Adaptive learning system for user-specific growth
  • Real-time personality adaptation algorithms

Technical Architecture

  • End-to-end encrypted data storage and transmission
  • Scalable WebSocket infrastructure for real-time chat
  • Multi-modal context understanding system
  • Advanced caching for sub-100ms response times

Key Features

  • Personalized AI companions with unique personalities
  • Dynamic conversation modeling and adaptation
  • Progress tracking and insight generation
  • Community-driven growth patterns

Impact & Metrics

  • 95% user satisfaction rate
  • Average session length of 15+ minutes
  • 50K+ meaningful interactions per day
  • 87% user retention after 30 days

MedVision - Medical Imaging AI

Advanced medical imaging analysis system using deep learning:

  • Custom CNN architecture with attention mechanisms
  • Transfer learning from ResNet50 and Vision Transformer
  • Achieved 96% accuracy on tumor detection
  • Implemented gradient-weighted class activation mapping
  • DICOM format support with metadata preservation

EdgeScale - Distributed Task Processing

High-performance edge computing platform leveraging Cloudflare's infrastructure:

  • Built with Cloudflare Workers and Durable Objects
  • D1 Database for persistent storage
  • R2 for large file handling
  • Processed 1M+ requests/day with 99.99% uptime

MultiLingua - Cross-lingual NLP Engine

Advanced NLP system for cross-lingual understanding:

  • Fine-tuned XLM-RoBERTa for 100+ languages
  • Implemented cross-lingual knowledge distillation
  • Built custom tokenization for Asian languages
  • Achieved SOTA results on XNLI benchmark

RetailVision - Retail Analytics System

Real-time computer vision system for retail analytics:

  • Custom YOLO-v5 model for product detection
  • Real-time tracking with DeepSORT algorithm
  • Edge deployment on NVIDIA Jetson devices
  • Achieved 94% accuracy in crowded scenes

CommitGenius - Contextual Commit Generator

Built an intelligent commit message generation system at BudSense that understands codebase context:

  • Developed semantic analysis system for understanding code changes across multiple files
  • Implemented codebase-aware RAG system for contextual understanding
  • Built git hooks integration for seamless developer workflow
  • Reduced average commit message writing time by 78%
  • Improved commit message consistency and clarity across team
  • Integrated with existing CI/CD pipeline for automated validation

Technical Implementation

  • LLM-powered analysis with custom fine-tuning for code understanding
  • Vector embeddings for efficient codebase context retrieval
  • Real-time diff analysis and semantic understanding
  • Conventional Commits specification compliance

Key Achievements

AI Innovation & Internal Tooling at BudSense

Advanced RAG Systems

  • Developed novel RAG optimization techniques for large-scale documentation
  • Implemented Semantic Prefix Reranking (SPR) for enhanced vector search accuracy
  • Achieved 87% improvement in retrieval relevance compared to baseline

AI-Powered Developer Tools

  • Built automated documentation generation system for massive codebase
  • Created AI-driven process diagram generator for complex workflows
  • Developed context-aware code generation system using up-to-date documentation
  • Implemented graph-based RAG for improved context understanding

Technical Impact

  • Reduced documentation time by 75% through AI automation
  • Improved developer onboarding time
  • Created scalable vector search infrastructure handling 100K+ documents

Startup Success

  • Founded and successfully exited a tech startup
  • Led technical strategy and architecture decisions
  • Built and managed high-performing engineering team
  • Developed scalable cloud infrastructure from ground up

System Design Excellence

  • Designed and implemented large-scale distributed systems
  • Created efficient vector search infrastructure for ML applications
  • Developed robust CI/CD pipelines for complex deployments
  • Implemented microservices architecture with high availability

Let's Connect

@sheroze_ajmalsheroze@codedrealities.com

Feel free to reach out for collaborations or just to say hello!