ANKITA GHOSH
GenAI · RAG · LLMs · NLP · VLM
🚀 Building Production-Ready AI Systems
01
RAG
LLM
NLP
0x1A
🎓 M.Sc. in Artificial Intelligence & Machine Learning (CGPA: 9.3) 💼 Ex-ML Intern @ Tech Mahindra Makers Lab | AI Intern @ Codec Technologies (Current)
🤖 GenAI & NLP Engineer — I build production-ready AI systems, not just models. 🌟 Contributor to vinta/awesome-python — the #10 most starred repo on GitHub ✅ 🏅 HackerRank — Active problem solver in Python, SQL & Algorithms
💡 Focus Areas:
- RAG Systems & LLM Pipelines
- Multilingual NLP (Indian Languages, 10,000+ dialect variations)
- Synthetic Data Generation (10× scaling)
- Document Intelligence & OCR
| # | Repo | Type | Description | Status |
|---|---|---|---|---|
| 1 | vinta/awesome-python | 🔀 PR | fix: Remove duplicate ruff entry from Code Linters section — Merged ✅ into #10 most starred GitHub repo! |
🟣 Merged |
| 2 | recodehive/Opensource-practice | 🔀 PR | fix: Resolve misplaced names & duplicate J section — Merged ✅ |
🟣 Merged |
| 3 | canonical/pycloudlib | 🔀 PR | docs: Update contributing guidelines |
🟡 Open |
| 4 | topoteretes/cognee | 🔀 PR | Add beginner-friendly example to documentation | 🟡 Open |
| 5 | recodehive/machine-learning-repos | 🐛 Issue | [Enhancement] Add Multilingual & Indian Language NLP Resources |
🟢 Raised |
| 6 | recodehive/Opensource-practice | 🐛 Issue | [Bug] Multiple names misplaced & duplicate section under letter J |
🟢 Raised |
| 7 | supabase/supabase-py | 💡 Issue | Suggestion to improve documentation clarity for beginners | 🟢 Raised |
| 8 | topoteretes/cognee | 💡 Issue | [Docs] Improve documentation with beginner-friendly knowledge graph example |
🟢 Raised |
| 9 | public-apis/public-apis | 🐛 Issue | [Bug] Multiple defunct/dead APIs still listed in README (FTX, AnonFiles, MetaWeather & more) |
🟢 Raised |
| 10 | Mohitha1406/text-emotion-classifier | 🐛 Issue | [Docs] Add dataset source and download instructions for emotion classification model |
🟢 Raised |
| 11 | shrutiii16/Traffic-Patterns | 🐛 Issue | [Docs] Add README with project overview, dataset description and usage instructions |
🟢 Raised |
| 12 | EpistasisLab/tpot | 🐛 Issue | [Docs] Fix typo: "feautures" should be "features" in README Tips section |
🟢 Raised |
| 13 | TheAlgorithms/Python | 💡 Issue | [Feature] Add Naive Bayes Text Classification Algorithm to machine_learning/ |
🟢 Raised |
| 14 | datasciencemasters/go | 🐛 Issue | [Docs] Update outdated book prices and mark freely available resources |
🟢 Raised |
| 15 | TheAlgorithms/Python | 🔍 Review | [Code Review] Reviewed PR #14665 — identified type hint bug in predict(), missing edge-case doctests for empty string input, validation order issue in fit(), and missing demo output in __main__ |
🟡 Reviewed |
🟢 Actively solving problems in Python · SQL · Problem Solving · Algorithms
- 🌍 Selected through the National Internship Portal (NCS) at Codec Technologies — a global IT & consultancy platform operating in 27+ countries
- 🤖 Working on applied AI/ML project assignments under Codec Technologies' Google for Education partner platform framework
- 🧠 Applying GenAI and deep learning skills to real-world business challenges
- 🚀 Building domain expertise in AI systems, production deployment, and cross-cultural AI problem solving
- 🎓 Completed internship as a Data Analyst and Python Developer at Huban Technologies LLP ▸ Developed an end-to-end OCR Invoice Data Extraction & Automation system using Python, EasyOCR, and machine learning techniques to digitize and process invoice documents automatically. ▸ Built an intelligent pipeline to extract structured fields (invoice number, date, vendor, line items, totals) from scanned and digital invoices using OCR and NLP-based post-processing. ▸ Automated data validation and export workflows, reducing manual data entry effort significantly and improving invoice processing accuracy through ML-based field classification. ▸ Integrated OpenCV for image preprocessing (deskewing, noise removal, binarization) to improve OCR accuracy on low-quality scanned invoice documents.
- 🌐 Built multilingual NLP pipelines handling 10,000+ dialect variations
- 🎙️ Developed ASR (Automatic Speech Recognition) systems with WER evaluation
- 🏷️ Designed automated annotation pipelines using NVIDIA NIM
- 🔄 Built real-time translation systems for low-resource Indian languages
- 📦 Built LLM-based Synthetic Data Generation pipelines achieving 10× dataset expansion
| Project | Description | Stack | Link |
|---|---|---|---|
| Netflix Content Dashboard | Interactive viz of 8,807 Netflix titles — Movies vs Shows, Country Map, Genres & Ratings | Tableau | |
| LLM Q&A Generation System | Auto-generates high-quality QA pairs from documents | LangChain, Gemini, RAG | |
| Intelligent RAG Pipeline | Multi-source document retrieval with semantic reranking | FAISS, LangGraph, HuggingFace | |
| Vision-Language Model | OCR + LLM fusion for document intelligence | VLM, Tesseract, GPT-4V | |
| Forecasting & ML Projects | Time-series forecasting with ensemble methods | scikit-learn, MLflow, Streamlit |
- Contribute to top open-source AI projects (LangChain, HuggingFace, etc.)
- Build and deploy scalable GenAI systems end-to-end
- Strengthen MLOps, system design & distributed training
- Earn HackerRank Gold badges in Python, SQL & Problem Solving 🏅
- Land a high-impact AI/ML engineering role 🚀
⭐ Building AI systems that actually work in the real world