Ali Faisal Khan

Gold Medalist AI/ML Engineer Specializing in Agentic AI and Production-Ready LLM Systems

Based in: Mardan, PakistanCurrently: AI Engineer, EY (Ernst & Young)

A GIKI Computer Science Gold Medalist with extensive experience in fine-tuning vision-language models and building agentic AI workflows for financial and healthcare sectors. He bridges the gap between research and production by leveraging Docker, CI/CD, and cloud-native tools to ship reliable AI-driven solutions.

Experience

AI Engineer

EY (Ernst & Young)Islamabad, PakistanworkOct 2026 – Present

Summary

Building financial AI use cases using LLM-based agents and agentic workflows for clients in the Middle East. Delivering client-facing AI solutions in a Big Four consulting environment.

What I did

Focused on Data & AI initiatives.

LLMAgentic AIAgentic workflows

Agentic AI Engineer

Starwall Inc.Ontario, Canadapart timeOct 2026 – Present

Starwall is an employee recognition and workplace culture platform with a mobile app and rewards marketplace. Building workflow automations and moving product development toward an AI-first approach, using AI agents with human oversight.

AI agentsWorkflow automation4-bit NF4 quantizationQLoRA

Research Intern

University of Europe for Applied SciencesPotsdam, GermanyinternshipJun 2025 – Aug 2025

Summary

Fine-tuned models on diverse image datasets to analyze model limitations; work resulted in three IEEE conference publications.

Results

Evaluated YOLOv8-Nano on 1,466 images for traffic-light detection under challenging conditions, achieving 80.4% mAP@0.5.
Developed a webcam-based interface using transfer learning with EfficientNetV2S to recognize 28 American Sign Language classes.
Compared exact and approximate N-Queens solvers across board sizes of 10–200, evaluating runtime, memory, and success rates.

Image datasetsModel fine-tuningYOLOv8-NanoEfficientNetV2STransfer LearningExperimental Design

Skills

Technical

LLMs

Agentic AI

Python

GitHub Actions (CI/CD)

Computer Vision (YOLOv8, EfficientNetV2)

ML Algorithms

LangChain

Docker

OCR (TrOCR)

VLM Fine-tuning

NLP

Supabase

FastAPI

Git

RAG

SQL

Linux

Django

PostgreSQL

Terraform

Node.js

C/C++

JavaScript

MongoDB

SQLite

AWS

Vision-Language Models (VLM)

Dataset Curation

Quantized Model Training

Experimental Design

Schema-based output structuring

Output validation

Ollama

Model quantization

Resource-efficient inference

4-bit NF4 quantization

QLoRA

Projects

meDDI - AI Prescription Analytics

Lead Developeracademic

Fine-tuned Qwen-3 VLM on handwritten prescriptions to extract structured medical data for downstream analysis. Detected drug-drug interactions by integrating multiple pharmaceutical databases; built analytics for pharmaceutical companies to support data-driven decisions. Collaborated with hospitals on AI-driven healthcare solutions.

Qwen-3 VLMPythonPharmaceutical DatabasesQwen VLMsOllama4-bit NF4 quantizationQLoRA

Visit the site

AI-Powered Resume Builder & Job Automation

personal

Automated resume generation and job-application workflows with LLMs (LangChain), ranking jobs by suitability to prioritize the best-fit applications.

LLMsLangChain

End-to-End DevOps Automation

personal

Built automated deployment pipelines using Linux, Docker, Terraform, and GitHub Actions CI/CD for faster, repeatable, error-free releases.

LinuxDockerTerraformGitHub ActionsCI/CD

Healthcare Document Intelligence Pipeline

personal

Developed a pipeline to extract structured data from handwritten prescriptions by fine-tuning vision-language models.

Qwen VLMQuantized model training

Traffic Light Detection Research

personal

Evaluated YOLOv8-Nano performance on 1,466 images for traffic-light detection in challenging conditions.

American Sign Language Recognition

personal

Used transfer learning with EfficientNetV2S to recognize 28 sign-language classes through a webcam-based interface.

EfficientNetV2STransfer Learning

N-Queens Solver Evaluation

personal

Comparative study of DFS, hill climbing, simulated annealing, and genetic algorithms across board sizes of 10–200.

DFSHill climbingSimulated annealingGenetic algorithms

Education

BS Computer Science

Ghulam Ishaq Khan Institute of Engineering Sciences & Technology (GIKI)Computer ScienceGPA 3.882022 – 2026

Thesis. meDDI - AI Prescription Analytics

Recognition

Awards

Gold Medalist

Ghulam Ishaq Khan Institute of Engineering Sciences & Technology (GIKI)2026

Awarded for academic excellence in BS Computer Science

Dean's Honor Roll

Ghulam Ishaq Khan Institute of Engineering Sciences & Technology (GIKI)

Achieved for all 8 semesters

Faculty Rank 1

Ghulam Ishaq Khan Institute of Engineering Sciences & Technology (GIKI)

Ranked 1st in the faculty

Batch Rank 4

Ghulam Ishaq Khan Institute of Engineering Sciences & Technology (GIKI)

Ranked 4th in the entire batch

Patents & publications

Benchmarking YOLOv8-Nano for Traffic Light Detection Under Challenging Conditions

publication2025

IEEE FIT Conference

Real-Time American Sign Language Recognition Using EfficientNetV2S: A Transfer Learning Approach with Webcam-Based Interface

publication2025

IEEE ICET Conference

Hire AliReach out about a role, a contract or a conversation.For recruiters

Ali's twin is AI, it can make mistakes.

More people on Kanvis

Browse more people Create your own Kanvis