Open to Associate AI Engineer Roles — RAG, LLM Agents & Applied GenAI
Associate AI/ML Engineer — Generative AI · RAG · LLM Agents · MLOps
I build production-ready RAG and agentic AI systems — from data pipelines and model evaluation to deployment. My work spans hybrid multimodal RAG, LLM-powered agents with LangChain and LangGraph, and shipping models as real services with FastAPI, Docker, and CI/CD.
— About
I'm an AI/ML Engineer with hands-on experience across the full GenAI stack — hybrid and multimodal RAG systems evaluated with RAGAS, agentic workflows built with LangChain, LangGraph, and n8n, and real-time voice AI on LiveKit. On the ML/DL side: supervised and unsupervised learning, and deep learning architectures including ANNs, CNNs, RNNs, LSTMs, and Transformers, with a strong NLP foundation. I deploy this work as real services using FastAPI, Docker, and GitHub Actions CI/CD. Built through applied training at Netsol Technologies; currently Co-Founder at Praxeon Solutions working on RAG systems, fine-tuning, and agentic workflows.
— Projects
Real-time Voice AI for Pakistani Students
Real-time voice agent acting as a career counselor for Pakistani students — checking eligibility, recommending degrees, and giving job-market outlook via live spoken conversation.
Multimodal Hybrid RAG for Research Papers
Multimodal hybrid RAG system for querying arXiv papers, ingesting PDF text, tables, and images with dense+sparse retrieval in Qdrant. A LangGraph agent orchestrates retrieval and generation, evaluated with RAGAS and shipped as a containerized FastAPI service with CI/CD.
Semantic Movie Recommendation API
Content-based movie recommender served as a live REST API, comparing TF-IDF and SBERT semantic search side by side. Models are cached on Hugging Face Hub and CI/CD auto-deploys to Hugging Face Spaces.
Human-in-the-Loop Multi-Agent Screening
Multi-agent resume screening pipeline in n8n. Skills matching, red-flag detection, and a hiring decision engine cross-validate results; ambiguous cases default to human review rather than auto-rejection.
Lightweight Transfer Learning for Tweet Sentiment
Tweet sentiment classifier for Negative, Neutral, and Positive labels using BERT Tiny with partial fine-tuning. Only the last two layers were trained to avoid overfitting on a 14,640-tweet dataset, reaching 74.45% accuracy while staying lightweight.
Agentic NLP-to-SQL with Intelligent Routing
Production-grade agentic NLP-to-SQL engine that translates plain English into structured queries, executes against an Aiven-hosted cloud database, and detects when a request needs visualization versus a clean data payload. The modular FastAPI backend is fully containerized.
Personalized Job Discovery with Semantic Matching
A job recommendation system that matches candidate skills and preferences to relevant opportunities using structured feature matching and NLP-based similarity. Designed to surface more useful roles than keyword-only search.
— Skills
Programming Languages
AI/ML Core
Generative AI & Agents
MLOps & Deployment
Mobile Development (Prior Experience)
Tools & Platforms
Familiar With
— Experience
July 2026 – Present · Lahore
Praxeon Solutions
Co-founded a platform focused on implementing published AI research and building applied systems. Currently working on hybrid multimodal RAG pipelines, RAGAS evaluation, and LLM fine-tuning — reproducing published methods, testing them against real-world constraints, and documenting findings along the way.
June 2026 · Lahore
NETSOL Technologies Pakistan
Built practical Generative AI applications, including a corporate website Q&A tool using Gemini Embedding v2 and Gemini 2.5 Lite, a PDF question-answering system backed by a FAISS vector database, and a natural-language-to-SQL tool with automatic data visualization.
February 2026 – June 2026 · Lahore
NETSOL Technologies Pakistan
Completed a 3.5-month intensive program covering Machine Learning, Deep Learning, Generative AI, RAG pipelines, and core statistics. Delivered technical presentations to senior leads and completed professional development training in resume building and interview preparation.
June 2025 – October 2025
Sky Logic Solutions Limited
Built a real-time voice chat room application using Flutter and Dart within a professional team. Developed responsive UI screens following MVC architecture, integrated and tested RESTful APIs, and implemented Zego Cloud SDK for real-time voice communication.
July 2025 – September 2025
Digital Empowerment Network
Completed a virtual internship with weekly task-based assignments in mobile development. Built full-featured apps including a wallpaper app and a recipe app, implementing Firebase Authentication and real-time database integration.
July 2025 – August 2025
SkillifyZone
Completed a 4-week internship building four Flutter mobile app projects. Implemented state management with GetX, integrated REST APIs, and followed MVC architecture for clean, modular code, delivering each project with documentation and demo recordings.
— Education
2021 – 2025
University of the Punjab
2019 – 2021
Qurban Degree College
2017 – 2019
Qurban Girls High School
— Credentials
Certifications
Introduction to Model Context Protocol
Anthropic (2026)
Model Context Protocol (MCP) · Generative AI Tools
Introduction to Subagents
Anthropic (2026)
AI Agents · AI Orchestration
Advanced Python
Simplilearn (2026)
Python · Advanced Concepts
Introduction to Prompt Engineering
Simplilearn (2026)
Prompt Engineering · Instructional Design
Object-Oriented Programming in Python
UniAthena (2026)
OOP Concepts · Python
Workshops
Part of GDG's #VibeKaregaPakistan series — a multi-city workshop track on building AI agents, from foundational concepts to real-world integration.
Intro to Transformers & Attention Mechanism
GDG PureLogics
Foundational concepts behind how LLMs actually work.
Agent Development Kit & AI Agents
GDG Daftarkhwan
Hands-on session on building AI agents using Google's Agent Development Kit.
Agent Integration in Android Apps
GDG Invozone
Applying agentic AI concepts inside real mobile app architecture.