Computer Science student at LUMS with a unique blend of deep technical research in vision-language models and award-winning leadership. As a Regional Head at AIESEC, I tripled department revenue and earned a National Excellence Award for top performance across Pakistan. Passionate about HCI and AI, I've built everything from RAG agents for legal tech to sophisticated fire detection systems, always focusing on the intersection of human needs and technical rigor.
Lahore, Pakistan
Experience
Feb 2025 – Feb 2026
WORK
Regional Head (Engagement)
AIESEC in Pakistan
Led cross-functional teams to plan and execute large-scale student engagement initiatives across regions. Tripled my department’s revenue, driving public engagement and conducting CSR initiatives. Won the national award for Excellence for the term 2025-2026.
•Led cross-functional teams to plan and execute large-scale student engagement initiatives across regions.
•Tripled my department’s revenue, driving public engagement and conducting CSR initiatives.
Key Achievements
→Tripled department revenue from 30k to 100k in one year by pivoting to social events, CSR, and strategic marketing of the 'exchange' product.
→Led the organization's creative hub, managing 30+ team members across the year (15 per 6-month term) to ideate and execute high-demand initiatives.
→Won the National Excellence Award for best performance across all portfolios in Pakistan (2025-2026).
→3.3x revenue growth (30k to 100k)
→Managed 30+ team members total
→National #1 ranking among all portfolios
2025 – Jun 2025
WORK
Engineering 100 Head TA
Lahore University of Management Sciences
Assisted in grading programming labs and assignments for first-year engineering students. Designed and assessed the end-of-semester Line Following Robot project, focusing on correctness, efficiency, and problem-solving approach.
•Assisted in grading programming labs and assignments for first-year engineering students.
•Designed and assessed the end-of-semester Line Following Robot project, focusing on correctness, efficiency, and problem-solving approach.
Key Achievements
→Led a team of 3 TAs to manage and grade 60+ students in the Engineering 100 course, overseeing 8 intensive 6-hour programming labs.
→Designed and assessed the end-of-semester Line Following Robot project, evaluating correctness, efficiency, and problem-solving approaches for 60 students.
→Managed 60+ students
→Led 3 TAs
→Oversaw 48+ hours of lab instruction
Programming
Project Portfolio
PERSONAL
Present
Campus Event Discovery - Software Engineering
Scrum Master
Gained hands-on experience with full-stack Android development by building a LUMS event management app with Stripe payment integration, QR-based ticketing, and ZXing-powered check-in scanning — learning how real-world payment pipelines and approval workflows are architected end-to-end. Developed a deeper understanding of asynchronous programming and Agile/Scrum practices by implementing an SOS geolocation alert system using parallelized Firebase reads, atomic batch writes, and FCM push notifications across a 5-person team.
Mashal - Job Matching Platform - Human Computer Interaction
UX Researcher & Designer
Learned to manage 4 distinct stakeholder groups through 18 semi-structured interviews, translating conflicting user needs into a cohesive 60+ screen bilingual Figma prototype — developing skills in user research, synthesis, and accessibility-first design for low-literacy audiences. Applied think-aloud protocol, Likert questionnaires, and screen recording across 20 usability test participants, achieving 4.27/5 mean usability and 100% task completion — learning how to close the gap between design intent and real user behaviour.
Learned to architect a complete Retrieval-Augmented Generation pipeline — from document ingestion and vector database indexing to semantic search and context-aware generation — gaining practical understanding of how grounding reduces LLM hallucinations in high-stakes domains. Developed skills in prompt engineering and retrieval optimization using LangChain, learning how retrieval logic and prompt design jointly determine response quality in domain-specific legal query systems.
Learned to design and evaluate vision-language hybrid architectures by fusing CLIP global embeddings into a YOLOv8n neck layer with identity initialization, gaining practical understanding of how pretrained representations can be adapted for domain-specific detection tasks. Developed rigorous experimental methodology through 12+ training runs and systematic ablations across lambda and alpha hyperparameters on a 21K-image benchmark, learning how to diagnose architectural failure modes through per-class performance analysis.
Pakiza - Transformer based Toxic Comment Detection - Natural Language Processing
NLP Developer
Learned to apply transfer learning to a low-resource, code-switched language by fine-tuning multilingual BERT embeddings on Roman Urdu data, gaining experience in adapting pretrained transformers to underrepresented linguistic contexts. Developed understanding of hierarchical classification pipelines by designing a two-stage architecture that first separates toxic from non-toxic content, then sub-classifies hate speech and racism as distinct categories.