Maliha Ehsan
CS Researcher & Developer · UMT Lahore
I study Computer Science at the University of Management and Technology, Pakistan. My work sits at the intersection of machine learning, explainable AI, satellite imagery analysis, and industrial fault diagnosis. I build things that are accurate and interpretable.
"The goal is to turn data into information, and information into insight."
— Carly FiorinaAbout
I'm a CS undergraduate at the University of Management and Technology (UMT), Lahore, actively working across deep learning, explainable AI, and geospatial ML. Currently a Research Assistant on two concurrent projects and a Flutter App Developer intern at Big Brains. Also an Experienced Student at Stanford's Code in Place program.
I care about building AI systems that don't just perform — they can be understood by the humans who deploy them. Interpretability-first, rigorous evaluation, readable code.
Tech Stack
Skills
ML & AI
Computer Vision
NLP & LLMs
Research & Product
Projects
Human Settlement Detection · Satellite Imagery
CNN segmentation pipeline detecting informal settlements from high-res satellite imagery across heterogeneous urban morphologies. Ongoing research at UMT — results being prepared for dissemination.
Adaptive Fault Diagnosis · Industrial Sensors
LightGBM pipeline optimized with EO-SSA (Elite Opposite Sparrow Search Algorithm) and CAMLLE dimensionality reduction for drift-aware fault classification in non-stationary industrial sensor data.
Urdu Fake News Detection System
Built in 48hrs at TECHVERSE AI Hackathon 2025. Accepts speech, English, and Roman Urdu — cross-references official news sources — returns result in under 4 seconds. Winner certificate issued by UMT Techverse.
Resume Matching · LLaMA Fine-Tuning
Fine-tuned LLaMA on recruitment data using parameter-efficient methods to match resumes by semantic relevance rather than keyword overlap, bridging the recruiter–candidate language gap.
Smart Van Scheduling · CSP Solver
AI scheduling system for vans, drivers, and delivery times using Backtracking, AC-3, MRV, Degree Heuristic, LCV, Forward Checking, Constraint Propagation, and Min-Conflicts. Produces conflict-free optimal schedules.
Voice Control Home Automation
End-to-end voice-controlled automation with Arduino Uno, Bluetooth HC-05, and an Android app. 90% recognition accuracy, sub-2-second response time, designed with hardware safety protocols.
Cyberattack Prediction · Network Traffic
ML pipeline for early-stage cyberattack prediction from network traffic using Random Forest and XGBoost with feature importance analysis to flag anomalous patterns before critical system compromise.
Research
Settlement Detection from Satellite Imagery
Automated detection of informal human settlements where conventional mapping fails across heterogeneous morphologies. CNN-based segmentation with geospatial preprocessing and benchmarking under distribution shift. Preparing for dissemination.
Adaptive & Explainable Fault Diagnosis in Industrial Sensor Systems
Drift-aware classification in high-dimensional industrial sensor data. LightGBM optimized via EO-SSA; CAMLLE dimensionality reduction preserves inter-class discriminability under distributional shift. Preparing for peer-reviewed submission.
Experience
Aug 2026 – Present
Flutter App Developer
Big Brains · Internship · Remote
Developing Flutter-based mobile applications as part of an internship team.
Mar 2026 – Present
Research Assistant — Satellite Imagery
UMT · Dr. Tayaba Anjum
- CNN-based segmentation for human settlement detection from high-resolution satellite data
- Data preprocessing pipelines, model training, and architectural benchmarking under distribution shift
Dec 2025 – Feb 2026
Research Assistant — Fault Diagnosis
UMT · Dr. Abdul Hameed
- LightGBM + EO-SSA hyperparameter optimization; CAMLLE supervised dimensionality reduction
- Drift-aware fault classification in high-dimensional non-stationary sensor data
Aug 2025 – Present
Graphic & eBook Designer
TeacherOn.com · Freelance · Remote
Creating visual identities and engaging layouts for diverse clients' digital publications.
Mar 2026 – Present
Experienced Student
Stanford University · Code in Place · Remote
Invited to Stanford's flagship distance-learning program — Python, computational thinking, software design. Full curriculum, certificate, and TA support.
Dec 2025 – Jan 2026
Youth Innovation Competitor
L'Oréal Brandstorm 2025 · Remote
Applied project management and SWOT analysis to develop a youth innovation proposal.
Licenses & Certifications
Product Management Job Simulation
Forage Academy · ID: 6a723ccb125ca4556ef7a6e1
Google AI Essentials
Google · ID: 499P3H3TF8KE
Automate Workflow with GitHub Actions
Microsoft Azure
Make the Sale: E-Commerce Stores
Google · ID: YE2EKK86HNHX
AI Fluency Framework & Foundations
Anthropic · ID: cus89gq5kuj5
Claude 101
Anthropic · ID: xmbooffw4ot9
AI Fluency for Students
Anthropic Education
Behavioural Science: Policy & Environment
ESCAP
Discover Data Analysis
Microsoft Azure
I&T Seminars 2026
Nestlé · ID: d7226e96-1609-4428-a347-92ce6229fb73
Entrepreneurship Training
Nestlé YEP Academy
MATLAB Onramp
MathWorks
Data Analytics Essentials
Cisco Networking Academy
TECHVERSE AI Hackathon 2025
UMT Techverse · Urdu Fake News Detection
Get Started with Power BI
Microsoft Azure
TensorFlow Developer Certificate
Introduction to AI
Education
University of Management and Technology (UMT)
B.S. Computer Science · Grade: A
AI, DSA, Linear Algebra, Software Engineering, ML, Computer Vision, NLP, HCI, OS, Deep Learning
Forman Christian College (A Chartered University)
Research Training Program
Citations Manager (Mendeley), Systematic Retrieval, Data Visualization with R
Lahore College for Women University
Pre-Engineering
Unique High School for Girls
Matriculation, Science · Grade: A+
Contact
Open to research collaborations, internship opportunities, and conversations about AI that matters.