AI engineer specialized in production ML systems and generative AI infrastructure.
Research depth in Bengali speech AI and low-resource NLP — one of the rarest intersections in the field.
Building systems that work at scale, not just in notebooks.
# Production ML · Generative AI · Agentic Systems · Bengali Speech AI · Low-Resource NLP
print("Build real systems. Ship to production. Compound the learning.")- Production-grade-RAG — Production RAG API built on FastAPI, LangChain, OpenAI, and Qdrant with retrieval evaluation and query monitoring
- Training Data Bot — Automated pipeline converting raw documents into quality-scored LLM fine-tuning datasets for QA, classification, and summarization tasks
- NeuroBridge — Middleware layer connecting LLMs to external systems and APIs for seamless production agentic integration
- GPT-2-Rebuild-nanoGPT — Step-by-step GPT-2 reconstruction from scratch, following Karpathy's methodology for deep architectural understanding
- GenAI-Roadmap-with-Notes-Using-LangChain — Structured GenAI learning path with hands-on LangChain implementations and architectural notes
- Agentic-AI-Roadmap-with-Notes-Using-LangGraph — End-to-end agentic AI curriculum with LangGraph implementations, from single agents to multi-agent orchestration
- 100 AI Machine Learning Deep Learning Projects — 100 production-oriented AI solutions spanning computer vision, NLP, and deep learning
- Prices_Predictor_System — End-to-end ML pipeline for price prediction with ZenML orchestration and MLflow experiment tracking
- ML-Algorithm-Library — Pure-Python implementations of core ML algorithms — built to understand, not just use
- ML Roadmap and Notes — Curated ML learning path with structured notes from fundamentals through advanced topics
- Book Recommender System — Collaborative and content-based recommendation engine using user reading history
- Laptop Price Predictor — Regression model predicting laptop prices across hardware configurations
- Stock Price Prediction — LSTM-based temporal model for stock price forecasting on historical data
- Machine_Learning_Housing_Corp — California housing value prediction benchmarked across multiple ML approaches
- Movie Recommender System — Personalized movie recommendations using content-based and collaborative filtering
- BUET-CSE-Fest-2026 ⭐ — Bengali ASR and speaker diarization systems built for BUET CSE Fest 2026 — private WER 0.37967 (Rank #28), private DER 0.28625 (Rank #19)
- Resume Screening — NLP pipeline for automated, bias-aware candidate evaluation
- Sentiment Analysis — Text sentiment classification deployed as an interactive Streamlit application
- Email Spam Classifier — High-accuracy spam detection using ML with NLP preprocessing
- Prompt Engineering — Curated resources for systematic prompt design and optimization
- Gesture Volume Control — Real-time hand gesture volume control using OpenCV and MediaPipe
- Posture Detection — Pose estimation system with configurable face overlay and real-time feedback
- Cat vs Dog Classification — CNN binary image classifier trained and benchmarked on the standard dataset
- QuantScope — Quantitative stock analysis platform combining fundamental and technical data for investment research
- Vendor-Performance-Analysis — Vendor metric analysis for procurement optimization and supply chain decision-making
- COVID-19 Global Impact Analysis — Statistical analysis of WHO global COVID-19 data across regions and time periods
- Olympic History Analysis — Visualization of 120 years of Olympic performance and national achievement trends
- Study Abroad Dataset — Interactive cost comparison for global study programs by country and institution
- PaperMap — Visual library transforming landmark AI research papers into structured, interactive educational experiences
- Toolly — AI-powered productivity platform powering toolly.tech
- Toolly_Studio — Streamlit app for professional product ad generation using Bria AI's image APIs
- Echo AI — Web-based interactive chatbot with a customizable interface
- Harvard Website — Clean, responsive portfolio site built on Harvard's visual design language
- 100 Simple Websites — Progressive front-end projects covering HTML, CSS, and JavaScript fundamentals
- Personal Website — Portfolio at adilshamim8.github.io showcasing projects and technical writing
- LearnEnglish — Interactive Bengali-to-English vocabulary platform covering the 3,000 most common English words
Kaggle Competition Solutions:AdilShamim8/Kaggle_Competitions
5 projects. Real systems. Zero hand-holding.
A hands-on curriculum for engineers who want to go from understanding LLMs to building production AI systems — step by step, project by project.
| Step | Project | What You Build | Core Skills |
|---|---|---|---|
| 101 | LLM Playground | An interactive environment to experiment with language models | LLM APIs Prompt Engineering Streaming |
| 102 | Customer Support Chatbot | A production-ready support bot with memory and context | RAG Memory Conversation Design |
| 103 | Ask the Web | An LLM agent with real-time web access and grounding | Tool Use Web Search Grounding |
| 104 | Deep Research | A multi-step autonomous research and synthesis system | Agentic Workflows Planning Synthesis |
| 105 | Multi-Modal AI | A system that processes vision, audio, and text together | Vision Audio Cross-modal Reasoning |
Each project builds on the last. Start at 101. Ship something real at every step.

