Hi, my name is
Humza Tareen.
AI researcher & engineer.
I combine rigorous research methodology with large-scale industry experience. My published work applies attention-driven deep learning and explainable AI to medical imaging, while my industry career spans training LLMs for Apple, Meta, and Bytedance, and building agentic AI systems at production scale. I'm actively seeking PhD opportunities to deepen my contributions to AI research.
Research
OptiGuard: Generalized, Attention-Driven & Explainable Glaucoma Classification
Glaucoma is a leading cause of irreversible blindness that demands early, accurate diagnosis. This work introduces OptiGuard, a generalized, attention-driven, and explainable system for glaucoma classification from retinal fundus images. The system employs a two-stage architecture: (1) optic disc and optic cup segmentation using Detectron2 with Mask R-CNN and a ResNet-50 backbone, validated with Average Precision metrics and computing diagnostic CDR and NRR values; (2) classification via EfficientNet-B0 with weighted sampling to handle class imbalance, trained on the SMDG-19 dataset with histogram equalization and CLAHE augmentation. Explainability is achieved through Grad-CAM, attention mechanisms, and LLM-generated reports, enabling clinicians to interpret and trust model predictions. The system achieved state-of-the-art results while maintaining clinical interpretability.
Research Interests
Explainable & Trustworthy AI
Developing models that are not only accurate but interpretable to domain experts. My published work uses Grad-CAM and attention-based explainability for clinical AI.
Computer Vision & Medical Imaging
Applying deep learning to healthcare challenges. Experience with fundus image analysis, optic disc/cup segmentation, and multi-dataset generalization.
Large Language Models
Hands-on experience training and evaluating LLMs for Apple, Meta, and Bytedance through RLHF, advanced reasoning benchmarks, and code interpreter development.
Agentic AI Systems
Architecting autonomous AI agents with RAG, tool-use, and multi-step reasoning. Building production-grade agentic backends with Python, FastAPI, and GCP.
Research & Industry Experience
Deep Learning Researcher @ National Center of Artificial Intelligence (NCAI)
- Authored a peer-reviewed paper (OptiGuard) published at the 47th IEEE EMBC on explainable AI for glaucoma detection, winning 1st Prize at the SEECS Open House.
- Designed and implemented a two-stage deep learning pipeline: (a) Detectron2 + Mask R-CNN (ResNet-50 backbone) for optic disc/cup segmentation, (b) EfficientNet-B0 classifier with weighted sampling and CLAHE augmentation on the SMDG-19 dataset.
- Integrated Grad-CAM and attention-based explainability with LLM-generated clinical reports, achieving state-of-the-art results with clinician-interpretable predictions.
- Conducted under the supervision of Dr. Muhammad Naseer Bajwa.
Software Engineer @ Turing
- Architected production microservices on GCP (Cloud Run, GKE, AlloyDB, Pub/Sub, Cloud Tasks) for an AI agent evaluation platform — including the core evaluation engine, automated LLM scoring service, RAG retrieval pipeline, and authentication gateway.
- Designed event-driven orchestration using Pub/Sub with dead-letter queues, Cloud Tasks, and Cloud Scheduler to coordinate multi-phase evaluation workflows with independent retry and zero silent failures.
- Led a platform-wide security audit identifying critical vulnerabilities (SQL injection, hardcoded credentials, unauthenticated endpoints) and implemented remediation using GCP Secret Manager and parameterized queries.
Pod Lead @ Turing
- Led a globally-distributed team of 10+ AI engineers across 3 continents, managing high-priority evaluation projects for key APAC clients.
- Developed quality control frameworks and mentored team members on advanced reasoning and evaluation techniques.
LLM Python Developer @ Turing
- Developed and evaluated domain-specific LLMs for Apple, Meta, and Bytedance, specializing in RLHF, advanced reasoning, and Code Interpreter development.
- Designed complex evaluation suites and benchmarking frameworks to ensure model performance, safety, and accuracy.
Software Engineer @ Royal Cyber Inc.
- Fine-tuned Llama models on AWS Bedrock for domain-specific financial services copilots.
- Co-developed the 'RC AI OPS' GenAI agent (Python, LangChain) automating error resolution, improving backend efficiency by 50%.
- Delivered GenAI integrations for Fortune 500 clients using enterprise iPaaS platforms (Mulesoft, Salesforce) and event-driven architectures (Kafka).
Technical Trainee & Intern @ Royal Cyber Inc.
- Built AI agents for enterprise Middleware Technologies (Mulesoft, IBM ACE/MQ, Salesforce OMS), reducing developer workloads by 60%.
- Earned Mulesoft Certified Developer (Level 1). Delivered API integration POCs with Kafka, Flink, and external services.
Software Engineer @ Bitnine Global Inc.
- Contributed to the core of Apache AGE, developing methods for the 'agtype' datatype enhancing PostgreSQL graph database integration.
- Engineered an enterprise automation framework for cross-platform product testing (Windows, CentOS, Red Hat, Debian, Ubuntu).
- Authored technical content for developer marketing.
Education & Certifications
National University of Sciences & Technology (NUST)
Bachelor of Science, Computer Science
October 2020 — June 2024
Final Year Project: OptiGuard (IEEE EMBC 2025) — Explainable deep learning for glaucoma classification. Advised by Dr. Muhammad Naseer Bajwa at NCAI.
Technical Skills
Research & ML
Programming & Tools
Data & Infrastructure
Certifications
Leadership & Community
NUST ACM Student Chapter
President (June 2023 — May 2024)
Rose through 4 leadership roles over 3 years and 7 months (Executive, Team Lead Liaison, Treasurer, President). As DevFest'21 Event Head in collaboration with GDG Islamabad, organized a 50+ member team and managed an audience of 600+.
Uns-E-Mahroom
Founder (June 2021 — November 2022)
Founded a community dedicated to empowering underprivileged communities. "Uns-E-Mahroom" translates to "Love for Deprived" — focused on helping people establish their own businesses and achieve self-sustainability.
Royal Cyber Inc.
Campus Ambassador (Nov 2021 — Jun 2023)
Organized seminars on full-stack e-commerce development and Salesforce CRM. Coordinated at NUST Career Fair, conducting 50+ applicant interviews for recruitment screening.