Hello! I'm Deeptesh Rout, a Computer Science and Engineering undergraduate at ITER, Siksha 'O' Anusandhan (SOA) University, graduating in 2027.

I'm interested in AI/ML engineering, machine learning research, backend development, and full-stack applications. I enjoy turning concepts into working projects, understanding the mathematics behind algorithms, and improving systems through experimentation.

• 🎓 Education: B.Tech in Computer Science and Engineering, ITER / SOA University
• 📊 Current CGPA: 9.23/10
• 💻 Programming: Python, Java, SQL
• 🧠 Interests: Machine Learning, Deep Learning, NLP, LLM applications, RAG, and backend engineering
• 🧩 Problem solving: 450+ LeetCode problems
• 🎯 Long-term direction: Become a strong AI/ML engineer and build reliable, useful AI systems
• 🌱 Currently improving: PyTorch, model evaluation, retrieval-augmented generation, APIs, and production engineering

I aim to build a portfolio that demonstrates real implementation, careful evaluation, clear documentation, and honest reporting—not just a list of technologies.

• Skills
• Projects
• Certifications and Learning
• Career Roadmap
• Research Interests
• Current Priorities
• GitHub Activity
• Connect

Skill-level note: This list includes tools I have used in coursework or projects and technologies I am actively learning. It does not imply expert-level proficiency in every item.

These are based on projects I've worked on. Add the corresponding repository links and update each status before publishing.

Focus: NLP, semantic similarity, skill extraction, Streamlit

• Extracts skills from resumes and job descriptions and compares the two.
• Uses embeddings/similarity to identify overlap and potential skill gaps.
• Uses O*NET skills data as a reference source and explores an LLM-assisted explanation workflow.
• Next improvements: add evaluation examples, improve extraction quality, and report measurable results.

Focus: Retrieval-augmented generation, code understanding, LLM applications

• Explores ways to answer questions about a codebase using structured context and retrieval-oriented techniques.
• Emphasizes useful answers grounded in the source code rather than unsupported guesses.
• Next improvements: create a reproducible benchmark, test on multiple repositories, and measure answer correctness.

Focus: Information retrieval, MiniLM, FAISS, BM25

• Experiments with combining semantic retrieval and lexical search for document discovery.
• Investigates how hybrid retrieval can improve access to relevant context.
• Next improvements: evaluate Recall@k, MRR, and answer grounding on a labelled query set.

Focus: Scikit-learn, classification, model evaluation

• Built an ensemble using soft voting with Logistic Regression, Random Forest, and XGBoost.
• Reported project metrics: 75.07% accuracy, 0.7952 ROC-AUC, and 0.5965 F1-score.
• The F1 score highlights why accuracy alone is not enough to judge a classifier.
• Next improvements: examine class imbalance, tune decision thresholds, and compare against simple baselines.

Focus: CNNs, PyTorch, OpenCV

• Built a convolutional neural network workflow for handwritten image recognition.
• Used image preprocessing to prepare inputs for the model.
• Next improvements: document the dataset split, add a confusion matrix, and compare model variants.

Focus: Generative AI, API integration, application development

• Explored a fitness-plan generation application using Vapi and Gemini.
• Focused on connecting a user-facing interface with an AI-generated plan.
• Next improvements: add clear safety boundaries, input validation, and reliable error handling.

Focus: APIs, frontend/backend integration, LLM workflows

• Builds practical experiments that connect application interfaces with language-model capabilities.
• Interested in making AI tools easier to use, test, and maintain.

Focus: Data preprocessing, feature engineering, Random Forest

• Practised an end-to-end tabular machine-learning workflow.
• Previously recorded a public score of approximately 0.78229; verify the competition page before treating this as a current achievement.

Important: Add repository links only after checking that the repositories are public and the READMEs accurately describe your contribution. Metrics above are project-reported results, not guarantees of general performance.

Learning and certificates I have recorded:

• Claude and Amazon Bedrock — Anthropic Academy
• Google 5-Day AI Agents Intensive
• Model Context Protocol (MCP) learning/certificate
• CNN Visualization learning/certificate
• Infosys Networking and Generative AI certificates

Other technical learning includes Python, NumPy, Matplotlib, SQL, data structures and algorithms, machine learning, and deep learning.

Add certificate URLs and issue dates when available. Only label a course as completed if you have completed it and can verify the certificate.

My goal is to grow into an AI/ML engineer who can build, evaluate, and deploy reliable systems. Research remains a possible long-term direction as I gain experience and identify meaningful problems.

flowchart TD
    A[Strong CS Foundations] --> B[DSA, OOP, DBMS, OS, Networks]
    B --> C[Machine Learning Fundamentals]
    C --> D[PyTorch and Deep Learning]
    D --> E[LLMs, Embeddings, RAG]
    E --> F[Evaluation and Reliable AI]
    F --> G[APIs, Docker, Deployment, Monitoring]
    G --> H[Production-Ready AI Projects]
    H --> I[AI/ML Engineering Roles]
    H --> J[Research Opportunities]

1. Strong fundamentals: algorithms, mathematics, software engineering, and databases.
2. Evidence of skill: complete projects with tests, documentation, and measurable evaluation.
3. Internships and collaboration: seek relevant opportunities and learn from code review.
4. AI/ML depth: develop practical competence with PyTorch, LLMs, RAG, and model evaluation.
5. Long-term growth: consider research or postgraduate study when it supports a clear career goal.

Areas I am interested in exploring as my experience grows:

• Trustworthy and evidence-grounded AI
• Retrieval-augmented generation (RAG) and information retrieval
• Large language model evaluation and hallucination reduction
• Natural language processing and multilingual systems
• Applied machine learning and model interpretability
• Efficient deployment of useful AI applications

These are interests, not claims of published research. Any future research project should have a specific question, baseline, evaluation protocol, limitations, and reproducible results.

• Strengthen Python and Java fundamentals.
• Keep practising DSA consistently.
• Improve statistics, linear algebra, and the mathematics behind ML.
• Build stronger PyTorch projects with documented evaluation.
• Improve existing projects instead of leaving many unfinished repositories.
• Learn API design, testing, Docker, and deployment.
• Prepare for AI/ML and software engineering internships.
• Keep my resume, GitHub repositories, and certificate links accurate and up to date.

1. Understand the fundamentals. Do not rely on copying code without understanding it.
2. Finish what I start. A complete, tested project is more useful than several unfinished ones.
3. Measure performance honestly. Report baselines, evaluation data, and limitations.
4. Build responsibly. Respect licences, privacy, and data permissions.
5. Document clearly. Every serious repository should explain setup, usage, tests, and known limitations.
6. Keep claims verifiable. Goals are not achievements, and a model score is meaningful only with context.
7. Learn continuously. Use mistakes, code reviews, and experiments to improve.

I'm happy to connect with students, developers, and researchers interested in AI/ML, NLP, LLM applications, RAG, and software engineering. I value thoughtful feedback, open-source collaboration, and opportunities to build systems that solve real problems.

• 💼 LinkedIn
• 💻 GitHub

If you find one of my projects useful, feel free to open an issue, suggest an improvement, or contribute through a pull request.