Open to opportunities

Kannaki kaliyaperumal

@kannakikaliyaperumal

Message

AI/NLP engineer building scalable, privacy-preserving ML pipelines for real-world applications.

Sweden

Message

What I'm looking for

I seek roles building end-to-end NLP/ML systems in teams that value privacy-first, production-grade deployments and measurable product impact.

I am an AI/NLP engineer who builds end-to-end systems that transform unstructured media into actionable structured data. I focus on scalable, language-agnostic pipelines for summarization, classification, and metadata extraction across audio, video, and text.

My work has reduced manual processing by 70% for a food-ordering system through a robust hybrid NER model, and I currently develop automated video transcription and summarization systems to improve post-production workflows. I deploy models in production using Docker, Kubernetes, and Azure DevOps.

My graduate research specialized in privacy-preserving federated learning, integrating differential privacy and secure multiparty computation to enable collaborative analysis of sensitive health data while maintaining strong privacy guarantees. I designed a Log-Normal DP mechanism and achieved a 77% model performance under a strict privacy budget.

I bring practical ML and software engineering skills across Python, PyTorch, TensorFlow, Hugging Face, and cloud/container platforms, and I value building trustworthy, privacy-first AI that delivers measurable operational impact.

Experience

Work history, roles, and key accomplishments

Current

AI/NLP Engineer Intern

Current

Strictly Post Production AB

Sep 2025 - Present (1 month)

Developed AI/NLP pipelines for multilingual transcription, summarization, classification, and metadata extraction from video, audio, and XML, automating media asset management and improving post-production workflow efficiency.

NLP Summarization Information Retrieval Metadata Python Machine Learning Pipeline Management

AI Engineer Intern

Niyata Infotech Pvt.

May 2025 - Aug 2025 (3 months)

Built a hybrid NER model combining BERT embeddings and Character-CNN, deployed with Docker/Kubernetes and Azure DevOps, improving domain-specific extraction accuracy and reducing manual order processing by 70%.

Named Entity Recognition PyTorch Hugging Face Docker Kubernetes Azure DevOps data preprocessing

AI Developer Intern

Realdini Studios

Jan 2025 - May 2025 (4 months)

Designed AI-driven NPC and enemy behaviors in Unreal Engine 5 and integrated language models for responsive NPC dialogue to enhance game immersion and environment interaction.

Unreal Engine 5 Game AI Behavior Design Language Models Python NPC Dialogue

Master's Thesis Researcher

Luleå University of Technology

Jun 2024 - Sep 2024 (3 months)

Developed a federated learning system combining differential privacy and secure multiparty computation, achieving 77% model performance under a privacy budget of ε=0.355 for decentralized health data analysis.

Federated Learning Differential Privacy Analysis Secure Aggregation Python Research

AI Research Intern

Annamalai University

Feb 2023 - May 2023 (3 months)

Designed a transformer-based image captioning model using InceptionV3 for feature extraction and LSTM for language generation, and deployed a Streamlit app for automated image/video captioning.

Computer Vision Transformers Deep Learning Streamlit