Hasnain Ali

Academic

Career

Verbatim text

Hasnain Ali
Cell: 923051508177 |  Email: ha3060763@gmail.com
LinkedIn: https://www.linkedin.com/in/hasnainali365/
Address: STREET 4 CANAL ROAD JHANG SADAR , Jhang , Pakistan
PROFESSIONAL PROFILE
Machine Learning Engineer with hands-on experience building, fine-tuning, and deploying machine learning models, including
modern large language models (LLMs). Strong foundation in Python, core machine learning algorithms, and NLP, with practical
experience implementing automated data preprocessing, training, and inference pipelines. Worked on model evaluation, prompt
optimization, and LLM fine-tuning for task-specific use cases, with exposure to deploying models through APIs and cloud-based environments.
EDUCATION
Software Engineering
School of Electrical Engineering and Computer Sciences , Islamabad , 2.75 (2026)
INTERNSHIP EXPERIENCE
Optical Network and Technologies Labs
16-Jun-2025 - 01-Sep-2025
As an ML Intern at ONT Lab , I worked on OTDR measurement data to estimate Quality of Transmission (QoT) and detect
impairments in optical networks. I implemented machine learning(transformers) models, conducted extensive data preprocessing on
large optical trace datasets(time series datasets), and applied explainable AI techniques, including SHAP-based feature attribution, to
interpret model decisions and identify key physical parameters affecting QoT.
FINAL YEAR PROJECT
Emotion Detection form Micro Expressions
Aimed to enable accurate recognition of genuine human emotions by analyzing facial micro-expressions using a lightweight,
production-ready deep learning pipeline. Developed an emotion recognition system that detects facial micro-expressions that are
brief, involuntary cues often missed by traditional models. Implemented a teacher–student architecture using self-supervised learning
for motion representation and knowledge distillation to train a lightweight, deployment-ready model.
Secure Encrypted Decision Tree Inference System
Built a client–server system using fully homomorphic encryption for decision tree inference to enable privacy‑preserving classification
where the server never sees client plaintext data
Smart Consent Form Desk Assistant
Built an LLM-powered extraction pipeline using Llama-3.1-8b-instant Model hosted by Groq and prompt-engineered JSON schemas
to automatically structure key clinical and consent fields from unstructured healthcare documents
TECHNICAL EXPERTISE
Large Language Models
Fine Tuning LLMs (LoRA, PEFT), Prompt Engineering, Embedding Generation, Semantic Search, Vector Databases (FAISS,
Chroma), RAG, Text Classification, Information Extraction, Sequence-to-Sequence Modeling, Generative AI, Open-Source LLMs
(LLaMA, Mistral, Phi)
Programming and Software Foundations
Python, PyTorch, TensorFlow, NumPy, Pandas, Scikit-learn, OOP, DSA, SQL, NoSQL, Database Management,, Version Control

AI enrichment