Curriculum
Vitae

B.Sc. in Computer Science

• Focus: Artificial Intelligence & Machine Learning
• Research Thesis: "C-MAT: A Modality-Invariant Transformer Framework for Neurodegenerative Disease Assessment" · Grade A
• CGPA: 3.72 / 4.00 (Merit Scholarship Recipient)
• Relevant Coursework: Algorithms, Data Structures, Applied ML, Pattern Recognition

Officer (Software Development)

• Engineered and shipped an internal KPI operations platform to production, spanning workflow discovery, full-stack delivery, security hardening, and release operations.
• Built with Next.js, Django REST Framework, PostgreSQL, Redis, Docker, capability-based authorization, auditability, and production-aligned test coverage.

Process Reengineering Intern · MD Secretariat / ICCD

• Designed ICCD WorkDesk to replace Excel and email task coordination for a 100-person compliance division.
• Defined role-aware ownership, delegation chains, audit history, reporting, attachments, and management visibility.

Computer Science Instructor

• Delivering comprehensive lectures on algorithms, data structures, and system architecture for O-Level curriculum.
• Developed custom learning materials to improve student engagement and programming fundamentals.

Junior Executive, Research & PM

• Contributed to foundational engineering workshops and research project planning.
• Collaborated on early-stage robotics prototypes and technical documentation.

Instructor (STEM)

• Conducted classes in Mathematics, Physics, and IT for students (Grades 5-10).
• Mentored students to improve conceptual understanding in core STEM subjects.

Project Lead & Researcher

• Designed a dual-stream transformer for AD, PD, and FTD assessment using MRI and resting-state EEG.
• Built ETL for 1,400+ records and combined cross-modal alignment, modality dropout, SHAP, and Attention Rollout.

Lead Researcher

• Developed "Knowledge Graph Separation" metric to quantify LLM embedding alignment with biomedical knowledge graphs.
• Fine-tuned Llama-3, Mistral, and Phi-3 using QLoRA; achieved 83.8% accuracy, outperforming BioMistral-7B.

Lead Researcher

• Architected a multi-task U-Net for simultaneous tumor segmentation and classification from ultrasound images.
• Achieved 85.4% Dice coefficient and 87.5% accuracy with statistical validation.

Independent Researcher

• Built a stacking ensemble model (XGBoost, LSTM) forecasting U.S. recessions with 96% accuracy and 100% recall.
• Applied SHAP analysis to validate economic indicators and model interpretability.

• Built a shared parsed document model for GFM, TeX math, chemistry, and multilingual content across preview, TXT, PDF, and DOCX.
• Added worker-based parsing, isolated browser exports, structured diagnostics, and explicit source-fallback handling.

• Built a full-stack platform transforming form data into pure LaTeX/PDF resumes.
• Engineered a live preview system with real-time compilation and cryptographic watermarking.

• Implemented a two-pass compiler with lexical analysis, AST generation, and 3-address code generation.
• Handled symbol table management, type checking, and scope resolution.