Khizar Sultan - Senior AI/ML Engineer

Generative AI, Agentic AI, LLMs, RAG & Chatbots.

Accomplished engineer with 6+ years experience and a Master’s in Data Science. Delivered 7+ industrial projects across telecom, healthcare, insurance, retail, and cybersecurity. 2× Microsoft‑certified (Azure AI & Data Fundamentals) and contributor to 20+ open‑source projects.

6+ yrs

Experience

Industrial Projects

2×

Microsoft Certified

20+

Open‑source Repos

Professional Summary

Accomplished AI/ML Engineer with 6+ years of experience and a Master’s in Data Science. Delivered 7+ industrial AI/ML projects across telecom, healthcare, insurance, retail, and cybersecurity. Proficient in Python, Generative AI, Agentic AI, Chatbots, and LLMs. 2× Microsoft‑certified in Azure AI and Data Fundamentals with 20+ open‑source projects on GitHub showcasing AI/ML expertise.

Work Experience

Sr. AI/ML Engineer — Gen AI

NETSOL Technologies · Pakistan · 09/2024 — Present

Built a conversational chatbot for a car lease company using Generative AI, LLMs, RAG, and LangChain; deployed with FastAPI, Docker, and AWS EC2 — increased car sales by 15% post‑deployment.
Developed a RAG‑based teacher assistant with LangChain, LangGraph, and Agentic AI to query submissions and auto‑email non‑submitters — saved ~5 hours/day per teacher.
Designed an MCP server using GoFastMCP, integrated with OpenAI and Claude Desktop; implemented intelligent tool calls and counter‑questioning for improved contextual accuracy.
Architected an API gateway on Databricks leveraging Mosaic AI to route user queries to the most relevant LLM automatically, reducing manual overhead.

AI/ML Engineer — Gen AI

SlashNext Inc · Pakistan · 07/2022 — 09/2024

Fine‑tuned LLMs (incl. Llama‑2) for domain‑specific data generation with chain‑of‑thought prompting — achieved 99% accuracy in malicious SMS generation.
Led BERT‑based SMS classification achieving 98% accuracy and contributing to a 40% decrease in cybersecurity attacks.
Implemented end‑to‑end CI/CD pipelines for text processing, model training, evaluation, and deployment — 25% reduction in train/test/deploy time.
Monitored production models and communicated progress via email, Confluence, reports, and presentations; collaborated with stakeholders to integrate new AI/NLP models into production systems.

Data Scientist — AI/ML Engineer

Retailo Technologies · Pakistan · 12/2021 — 06/2022

Built a recommendation engine (Collaborative Filtering) contributing to $1M revenue.
Extracted/cleaned sales data from AWS Redshift with Python; engineered features using Pandas/NumPy; improved prediction accuracy with RF/GBMs.
Optimized storage with compression and indexing — 40% faster access times; added KPI tracking to quantify impact.

Associate Data Scientist — AI/ML Engineer

Addo.ai · Pakistan · 10/2020 — 12/2021

Delivered solutions for international clients (Singtel, ComfortDelGro) — 15% efficiency boost.
Participated in daily scrums — reduced project turnaround by 10%.
Used Python/easyocr/numpy/pandas/matplotlib/sklearn — 25% reduction in data processing time.
Trained/validated ML models and deployed via Docker/Kubernetes — 20% faster deployments and 30% higher model accuracy.

Industry Projects

Enterprise Conversational RAG Chatbot

LLMs, RAG, LangChain, LangGraph, Agentic AI · FastAPI · Docker · AWS EC2

Objective: Handle complex business queries and automate workflows.

Approach: Context‑aware interactions; task automation (emailing non‑submitters, retrieving student data); scalable deployment.

Impact: +30% user engagement and reduced manual effort across ops.

Car Lease Sales Chatbot

Generative AI, LLMs, RAG · LangChain · FastAPI · Docker · AWS

Objective: Drive sales with personalized conversational assistance.

Approach: Retrieval‑augmented grounding, lead qualification, and quote assistance.

Impact: +15% car sales post‑deployment.

Teacher Assistant RAG Agent

LangChain · LangGraph · Agentic AI · Email Automation

Objective: Query and manage student submissions automatically.

Approach: Agent tools for data lookup and batch emailing; guardrails and evals.

Impact: Saves ~5 hours/day per teacher.

Malicious SMS Classification (NLP)

BERT · Transformers · Python · CI/CD

Objective: Detect malicious SMS to enhance mobile security.

Approach: BERT fine‑tuning with domain data and CoT prompts.

Impact: 98% accuracy; 40% fewer cyber attacks.

Malicious Email Classification

TF‑IDF / Word2Vec · Anomaly Detection · Random Forest

Objective: Detect and classify malicious vs. benign emails.

Approach: Feature extraction from subject/body/sender; supervised + anomaly detection.

Impact: 95% accuracy; 40% reduction in malicious emails reaching users.

MCP Server for Intelligent Tool Orchestration

GoFastMCP · OpenAI · Claude Desktop

Objective: Improve multi‑model AI with dynamic tool calling.

Approach: Counter‑questioning & context routing for accurate tool selection.

Impact: Higher response accuracy and reduced manual tool invocation.

OCR Identity Card Extraction

easyocr · Python

Objective: Extract name/DOB/address/ID fields from ID cards.

Approach: Image preprocessing + deep‑learning OCR pipeline.

Impact: 40% faster verification for gov/banking/travel flows.

Retail Recommendation Engine

Collaborative Filtering · AWS Redshift · Python

Objective: Personalize product recommendations.

Approach: Data ingestion/feature engineering; CF modeling; KPI tracking.

Impact: Contributed to $1M revenue.

Education

M.Sc. Data Science

University of Management and Technology — Pakistan (07/2021 — 07/2023)

Major Courses: AI, NLP, Data Science, Machine Learning, Deep Learning, Computer Vision

B.Sc. Computer Science

National University of Computer and Emerging Sciences — Pakistan (07/2016 — 07/2020)

Major Courses: DSA, Programming, Statistics & Probability, Calculus, Linear Algebra, Data Mining