Chirag Sharma

Computer Science & Engineering | B.Tech, MSIT Delhi 📍 New Delhi, India

Motivated 3rd-year B.Tech CSE student with real-world experience in Machine Learning, NLP, and developing data-powered applications. Strong command of Python, SQL, PyTorch and Scikit-learn.

Experienced in architecting end-to-end AI workflows, optimizing data pipelines, and building analytics dashboards with a strong focus on scalability, reliability, and production deployment.

💼 Experience

Edunet Foundation

Artificial Intelligence & Data Analytics Intern

Developed a Carbon Emission Forecasting tool using RandomForestRegressor trained on historical data (1990–2008) to predict 20 years of future emissions (2009–2028).
Achieved a mean cross-validated R² score of 0.986 after hyperparameter tuning.
Demonstrated how predictive analytics can drive environmental monitoring and awareness generation.
Contributed to long-term climate strategy formulation using data-driven insights.

🌐 Open Source Contributions

Matplotlib

Added Ctrl-based 3D rotation snapping with 5-degree increments and introduced the axes3d.snap_rotation rcParam for configurable angle snapping.
Fixed missing xlim_changed / ylim_changed callbacks when axis limits expand via set_xticks / set_yticks, ensuring consistent behavior.
Investigated a DST-related bug in AutoDateLocator and proposed a UTC normalization approach to resolve tick spacing issues caused by repeated timestamps.
Improved documentation for SecondaryAxes limit behaviour,clarifying that set_xlim / set_ylim on secondary axes are ineffective and suggesting better API usage.

pvlib-python

Added usage examples for run_model_from_poa and run_model_from_effective_irradiance in pvlib ModelChain, including single-array and multi-array workflows.

ArviZ

Implemented __array__ protocol support in convert_to_datatree, improving compatibility with array-like libraries.

PyMC

Fixed a crash in vectorize_over_posterior when using distributions like ZeroSumNormal by removing the explicit dtype argument.
Rejected “/” in variable names via _validate_name(), preventing conflicts with Xarray/DataTree.
Added validation in Model.add_coord and Model.add_named_variable to prevent coordinate and variable name collisions, ensuring models fail early instead of causing silent downstream issues

🧠 Projects

Flipkart Sentiment Analysis

Fine-tuned DistilBERT on 205K+ Flipkart reviews achieving 96.27% validation accuracy.
Implemented fp16 mixed precision training and class imbalance handling with random oversampling.
Built and deployed real-time inference pipeline with confidence scoring using PyTorch and HuggingFace Transformers.

Tech: DistilBERT · PyTorch · HuggingFace Transformers · NLP

RAG File QA System

Built a RAG-based Code Assistant enabling semantic Q&A, debugging, and code generation for Python files.
Implemented FAISS vector search with Sentence Transformers and Tree-sitter AST parsing.
Used Mistral-7B as the backbone LLM for context-aware responses.

Tech: RAG · FAISS · Sentence Transformers · Mistral-7B · AST

Steel Fault Detection

Developed and deployed an end-to-end ML pipeline for steel fault detection using high-dimensional sensor data.
Applied feature engineering, dimensionality reduction, and imbalance handling techniques.
Benchmarked XGBoost, SVM, and Random Forest for optimal fault classification.

Tech: XGBoost · SVM · Random Forest · Scikit-learn

🎓 Education

Maharaja Surajmal Institute of Technology, GGSIPU

B.Tech in Computer Science & Engineering | 2023 – 2027 | CGPA:9.52

DAV Public School

Grade 12 | 2022 – 2023 | Percentage:90.6