Shaharyar T.

I reduced false-positive alerts by 82% for a power plant client, took an NLP system from 7% to 67% accuracy while cutting 700 manual rules down to 5, and dropped daily complaints by 60% for a telecom with over a million customers. That's the kind of AI and machine learning I build: work that moves a real number, not a demo that dies after the pitch. I build: • Custom ML models and predictive analytics (XGBoost, Scikit-learn, anomaly detection, demand and sales forecasting) • NLP and LLM systems (BERT fine-tuning, sentiment analysis, entity recognition, text summarization) • RAG chatbots and AI agents (LangChain, LlamaIndex, OpenAI/ChatGPT, Neo4j knowledge graphs, vector databases like Milvus and Pinecone) • Computer vision applications (image processing, colorization, OpenCV) • Data engineering pipelines (ETL, Airflow, Kafka, AWS Lambda, S3, Step Functions) • Web scraping and data collection at scale (Selenium, Scrapy, Apify) • Interactive dashboards and reporting tools (Plotly, Dash, Bokeh) Delivered this work across healthcare, insurance, telecom, energy, retail, e-commerce, restaurants, recruitment, social media, automotive, education, and maritime shipping, for clients in the US, UK, Canada, UAE, Australia, Europe, and Brazil. Stack: Python, Django, Flask, FastAPI, TensorFlow, Keras, Hugging Face Transformers, OpenAI API, LangChain, Docker, AWS, PostgreSQL, MySQL, Pandas, NumPy. I scope clearly before writing code, give honest feedback if your spec has gaps, and send regular updates instead of going dark until the deadline. Message me with what you're building, and I'll tell you straight whether I're the right fit.