Help & Supportcareers@gravityer.com
AI Engineer with less than a year in Multi-Modal Deepfake Detection & Machine Learning
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Highly motivated Computer Science student with a strong foundation in AI/ML, actively engaged in cutting-edge research and software development. Experienced in building multimodal deepfake detection frameworks, developing enterprise banking systems, and contributing to open-source AI projects. Eager to apply expertise in deep learning, natural language processing, and distributed systems to innovative challenges.
University of Engineering and Technology, Lahore
Bachelor of Science · Computer Science
September 1, 2023 – June 1, 2027
University of Engineering and Technology (UET), Lahore
Undergraduate Researcher - Multi Modal Deepfake Detection
January 1, 2026 – Present
Lahore, Punjab, Pakistan
National Bank of Pakistan (NBP) – IT Center & Digital Banking, Lahore
Software Engineer Intern
August 1, 2024 – September 1, 2024
Lahore, Punjab, Pakistan
AgentForce — AI-Powered Developer Automation Platform
June 26, 2026 – Present
Built an event-driven developer platform that automates testing, documentation, and collaboration workflows. Implemented a webhook-driven testing pipeline that automatically triggers on GitHub pushes, extracts repository code, generates test cases using Groq LLaMA, executes them inside isolated Docker environments, and delivers summarized test reports to developers. Integrated GitHub, Jira, Slack, and Notion to generate documentation and collaboration insights automatically. Developed an MCP-based AI assistant for secure tool orchestration.
View ProjectMLTrainer — Natural Language to Deployed ML Models
June 26, 2026 – Present
Built an autonomous multi-agent system that converts natural language requirements into complete machine learning pipelines and deployable Hugging Face models. Implemented specialized agents for dataset preparation, preprocessing, architecture selection, training, deployment, and monitoring. Developed a tool-calling framework enabling agents to interact with external services, including Hugging Face, Kaggle, local execution environments, and deployment workflows. Added checkpointing, SQLite-backed persistence, intelligent retry mechanisms, and multi-provider LLM support through LiteLLM. Supports computer vision, NLP, tabular learning, clustering, forecasting, and LLM fine-tuning workflows.
View ProjectDeepfake Detection Using Hybrid EfficientNet-ViT Architecture
June 26, 2026 – Present
Developed a deepfake detection system combining EfficientNet-B0 and Vision Transformers (ViT) to capture both local facial artifacts and long-range visual dependencies. Built preprocessing pipelines using FaceForensics++ and Celeb-DF datasets, including frame extraction, MTCNN-based face detection, augmentation, and inference. Achieved 88.64% classification accuracy and implemented end-to-end video classification and Hugging Face deployment workflows.
View ProjectLoopFlow
June 26, 2026 – Present
Built LoopFlow, an open-source TypeScript framework for designing autonomous agent loops around Claude Code, enabling developers to move from prompt engineering to loop engineering. Designed a YAML-based workflow system where developers define reusable loops that iteratively guide agents toward a goal through specialized steps, reviewer feedback, memory, budgets, and completion criteria. Implemented multi-step agent orchestration, persistent memory, independent review gates, cost controls, Git worktree isolation, and session replay support for reliable long-running agent workflows. Developed starter loops for automated test fixing, technical debt auditing, and documentation synchronization, reducing repetitive engineering effort through autonomous execution.
View ProjectFIBpeTokenizer — Rust-Based BPE Tokenizer for LLMs
June 26, 2026 – Present
Built a Byte Pair Encoding (BPE) tokenizer library from scratch in Rust with Python bindings through PyO3. Implemented vocabulary construction, token merging, encoding, decoding, model persistence, and configurable pre-tokenization strategies. Leveraged Rayon for parallelized training and Aho-Corasick for efficient token processing. Published on PyPI with 17,000+ downloads and benchmarked against Hugging Face tokenizers.
View ProjectFrame-Level Multimodal Deepfake Detection Using Bidirectional Cross-Attention and Temporal Transformer
Unknown
June 1, 2026 – Present
Cultural Fit Analysis
The candidate's portfolio showcases a strong passion for AI and machine learning, with a focus on building innovative, autonomous systems. The open-source contributions and diverse project work suggest a collaborative mindset and a willingness to engage with the broader technical community. The projects align well with a culture that values innovation, deep technical expertise, and practical application of AI research. The candidate's academic background and research experience further reinforce a fit for a technically rigorous and research-oriented environment.
Soft Skills & Operational Fit
The candidate demonstrates strong initiative and self-direction through numerous personal projects and open-source contributions. The detailed project descriptions suggest good communication of technical concepts. The breadth of technologies and project types indicates adaptability and a proactive learning attitude. The research experience also points to strong analytical and problem-solving skills.