Maitrayee

AI Engineer with 3+ years of hands-on experience building production-grade LLM applications, Retrieval-Augmented Generation (RAG) pipelines, and Agentic AI systems. Proficient in Python, LangChain, OpenAI API (GPT-4/GPT-4o), Anthropic Claude API, vector databases (ChromaDB, FAISS), and modern ML frameworks. Delivered end-to-end AI products including a multi-stage RAG research agent, an NLP-powered e-learning platform, and deep reinforcement learning agents.

Proven ability to architect, build, and ship AI solutions independently in fully remote and async environments. Actively seeking remote AI/ML engineering roles with US/EU companies.

Freelance AI/ML Engineer at Self-Employed (2023-01 – Present)

• Architected and shipped a production-grade AI Research Agent powered by a 5-stage RAG pipeline (web search, document parsing, semantic chunking, ChromaDB vector storage, GPT-4 report generation); built an automated claim-verification layer that cross-validates every LLM output against retrieved source documents, cutting hallucinated or unsupported claims by ~40% and meaningfully improving response trustworthiness for end users.
• Designed and delivered a full-stack Smart E-Learning LMS in Django supporting 3 distinct user roles (student, instructor, admin); built an explainable AI recommendation engine combining collaborative and content-based filtering on 5 behavioral signals (progress, quiz scores, wishlist, enrollment, engagement), integrated NLP-powered quiz generation using Hugging Face Transformers, automated plagiarism detection, and real-time learner analytics dashboards.
• Engineered production NLP pipelines using Hugging Face Transformers, SentenceTransformers, and LangChain for semantic search, text classification, and named entity recognition (NER); served models via FastAPI REST endpoints, achieving low-latency inference suitable for real-time applications.
• Trained a Deep Q-Network (DQN) reinforcement learning agent on OpenAI Gym Lunar Lander using TensorFlow; implemented experience replay, target networks, and epsilon-greedy exploration — converging to a stable landing policy after training and deploying the agent as a live interactive public web demo.
• Re-implemented 6 ML algorithms from first principles (Linear Regression, Logistic Regression, Decision Trees, K-Means Clustering, Anomaly Detection, Collaborative Filtering); benchmarked each against Scikit-learn baselines on standard datasets, validating correctness across loss functions, gradient descent convergence, and key metrics including F1, AUC, and RMSE.
• Mentored 5+ international remote students in Python, machine learning, and applied mathematics through structured async learning plans; consistently improved student grades from ~70% to 80-90% by producing written deep-dives on gradient descent, statistical inference, bias-variance tradeoff, and model evaluation strategies.

Data Science & AI Skills Intern at Microsoft (2025-07 – 2025-07)

• Completed Azure AI Fundamentals training; built NLP workflows using Azure AI Language for text analytics and sentiment analysis in a fully remote, async program
• Worked through generative AI workflows in Azure Machine Learning and completed Power BI data visualisation training.

B.E. in Information Technology – SAL College of Engineering, GTU

Diploma in Information Technology – SAL College of Engineering, GTU