AI Engineer with 1+ years in Deep Learning, LLMs & Generative Models
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
CS graduate (Dean's Honor List, FAST-NUCES) specializing in deep learning and applied machine learning. Built transformer encoder-decoders from scratch in PyTorch, fine-tuned LLMS (BERT, GPT-2, T5) with LoRA/PEFT on multi-GPU setups, implemented diffusion models and Vision Transformers from base PyTorch, and engineered hybrid-retrieval RAG pipelines evaluated with RAGAS. Hands-on with multimodal fusion and SHAP-based XAI. Published technical writing on Medium.
National University of Computer and Emerging Sciences (FAST-NUCES)
BS · Computer Science
August 1, 2022 – July 1, 2026
FynkTech Pvt Ltd
Full-Stack Engineer Intern
October 1, 2025 – January 1, 2026
Lahore, Punjab, Pakistan
FAST-NUCES
Research Assistant
January 1, 2025 – Present
Faisalabad, Punjab, Pakistan
FAST-NUCES
Teaching Assistant
January 1, 2025 – Present
Faisalabad, Punjab, Pakistan
Vision Language Model Fine-Tuning: Qwen2-VL with QLORA for Document-to-Markdown
March 1, 2026 – June 1, 2026
• Fine-tuned pretrained Qwen2-VL-2B-Instruct on the Nougat document dataset using QLoRA (4-bit NF4 quantization, frozen base weights, LoRA adapters rank 8-16) for document-image-to-Markdown generation; preprocessed image-text pairs into ChatML, trained on Kaggle T4×2 with gradient accumulation, deployed a Gradio app for real-time inference.
Denoising Diffusion Probabilistic Model (DDPM) from Scratch
March 1, 2026 – June 1, 2026
• Full DDPM pipeline from base PyTorch (forward noising, reverse denoising, sinusoidal time-step embeddings, residual U-Net 64-128-256) for high-resolution image generation on CelebA-HQ, FFHQ, and WikiArt at 128×128 and 256×256; trained on Kaggle T4×2 with AMP, evaluated via PSNR and SSIM, deployed an interactive Gradio app.
View ProjectMasked Autoencoder (MAE): Self-Supervised Vision Transformer Pre-Training
March 1, 2026 – June 1, 2026
• Asymmetric Transformer MAE from base PyTorch: ViT-Base (B/16) encoder (12 layers, 768 hidden dim) processing only 25% visible patches paired with a ViT-Small (S/16) decoder reconstructing 75% masked content on TinyImageNet at 224×224; trained on Kaggle T4×2 with AdamW + cosine LR + AMP, deployed a Gradio app with configurable masking ratios.
NeuroVerse: Multimodal Explainable AI for Clinical Diagnostics (FYP)
September 1, 2025 – June 1, 2026
• Multimodal XAI platform for Alzheimer's and Parkinson's disease detection that fuses speech, cognitive, facial, and motor biomarkers across 15,000+ samples; targeting 87% accuracy (14% above single-modality baselines) with SHAP feature attribution, deployed via Flutter + Supabase.
LLM Fine-Tuning Suite: BERT, GPT-2, and T5 across Three Architectures
August 1, 2025 – September 1, 2025
• Fine-tuned BERT (encoder-only) for 3-class sentiment classification, GPT-2 (decoder-only) for pseudo-code to Python on SPOC, and T5 (encoder-decoder) for summarization on CNN/DailyMail; applied LORA/PEFT to cut trainable parameters by approximately 90% versus full fine-tuning, benchmarked on a 4-GPU PyTorch DDP setup, and evaluated with F1, BLEU, CodeBLEU, and ROUGE-1/2/L.
Clinical RAG System: Diagnostic Reasoning on MIMIC-IV (DiReCT)
June 1, 2025 – August 1, 2025
• Hybrid retrieval pipeline (BM25 + dense FAISS) over 50,000+ MIMIC-IV-Ext clinical records with Groq LLaMA 3.3 via LangChain, achieving 89% precision@5 across 200+ test queries; evaluated faithfulness and answer relevancy with RAGAS, deployed Streamlit frontend with a published Medium writeup.
IEEE Project Exhibition
IEEE
January 1, 2023 – Present
Cultural Fit Analysis
The candidate's academic background and diverse project portfolio in cutting-edge AI/ML topics align well with an innovative and research-driven culture. The involvement in an IEEE Project Exhibition and Dean's Honor List indicates a drive for excellence and continuous learning. The full-stack internship shows adaptability beyond core AI, suggesting a willingness to engage with broader engineering challenges. However, the lack of extensive industry experience means their adaptability to specific corporate cultures and rapid product development cycles is yet to be fully proven.
Soft Skills & Operational Fit
The candidate's experience as a Research Assistant and Teaching Assistant at FAST-NUCES suggests an ability to work collaboratively, explain complex concepts, and manage responsibilities. The Full-Stack Engineer Intern role indicates exposure to Agile methodologies, API design, and real-time data synchronization, which are valuable for operational fit. However, the primary focus on academic projects means real-world operational experience in a fast-paced industry setting is limited.