Engineering robust AI solutions that scale from edge devices to the cloud.
Core competencies
AI & Machine Learning
Custom model development end-to-end, from computer vision solutions using Depth Anything V2 to production-quantized LLMs like Gemma 4.
Data Science & Analytics
Precision sales forecasting and predictive analytics built on robust data pipelines. From raw ingestion to actionable dashboards.
Agentic Workflows
Deep expertise in multi-agent orchestration using LangChain, LangGraph, and CrewAI. Building autonomous systems that reason and execute.
EYE-CUE
An on-device AI assistant fusing monocular depth estimation with semantic segmentation, guided by a quantized Gemma 4 running entirely on edge hardware. No cloud. No latency.
Dual-Stream Perception
Depth + segmentation running in parallel. Distance tells you how far; semantics tells you what.
Obstacle Fusion
Filters walkable surfaces and focuses guidance only on hazards that actually matter.
Path Planning
Stable, human-friendly directions without jitter from hard thresholds.
Multimodal Assistant
On-device LLM. Describes surroundings for blind or low-vision users.
Voice Navigation
Spoken guidance so the user can keep moving hands-free.
SOS Alert Module
Sends an emergency alert with photo and location when triggered.
Selected Works
Showcasing a few robust builds across AI, CV, and Data.
Focus Tracker
A lightweight overlay widget using computer vision to track face and emotional state during work. Shows live attention and engagement levels with personalized AI coaching messages based on mood.
SISA Machine Unlearning
Implemented the Sharded, Isolated, Sliced, and Aggregated (SISA) framework for machine unlearning (GDPR "Right to be Forgotten"). Enables efficient dataset deletion requests by partitioning training data and caching checkpoints, reducing retraining costs by 90% without rebuilding models from scratch.
Financial RAG Agent
Designed and deployed an enterprise RAG pipeline for a prominent Italian advisory firm, automating the analysis of complex financial documents. Engineered semantic chunking and hybrid vector search to eliminate a 24–48 hour manual data-auditing bottleneck—slashing retrieval times to under 2 minutes with near-zero hallucinations.
SPOTIFY RECOMMENDATION SYSTEM
MFCC feature extraction from audio files with metadata enrichment. A collaborative filtering music recommendation system using Apache Spark's ALS algorithm.
A word from people I've worked with
Global collaborations across AI and data science.
"Muneeb truly impressed me with their AI Development skills! His work not only demonstrated exceptional documentation and professionalism, but also went ABOVE and beyond expectations. Collaborating with them was a breeze, thanks to their politeness, quick responsiveness, and proactive communication."
"Muneeb is super fast, great in building ai systems that are functional to RAG LLM building. If you need a RAG LLM, consider hiring Muneeb!"
"Second project with Muneeb and he has been wonderful in delivering in a really short time something that is working really well! Very happy!"
"Very good expert ! 5 stars on 5 stars"
"Exceptional"