Sai Prasanth

AI and ML Engineer

Bengaluru, Karnataka 85 · Excellent

About

AI/ML Engineer with hands-on experience in Machine Learning, Generative AI, and RAG-based applications. Skilled in building predictive models, NLP pipelines, and real-time AI systems using Python, TensorFlow, LangChain, and Scikit-learn. Experienced in developing practical AI solutions including fraud detection, computer vision systems, and LLM-powered applications. Passionate about solving real-world problems through AI, continuously learning emerging technologies, and building scalable intelligent systems.

Skills & Expertise (40)

Python Expert

9.0/10

Years Exp

Machine Learning Expert

9.0/10

Years Exp

RAG pipeline Advanced

8.5/10

Years Exp

YOLOv8 Advanced

8.5/10

Years Exp

NLP Advanced

8.5/10

Years Exp

LightGBM Advanced

8.0/10

Years Exp

Random Forest Advanced

8.0/10

Years Exp

XGBoost Advanced

8.0/10

Years Exp

TensorFlow Advanced

8.0/10

Years Exp

OpenCV Advanced

8.0/10

Years Exp

Prompt Engineering Advanced

8.0/10

Years Exp

Computer Vision Advanced

8.0/10

Years Exp

React.js Advanced

7.5/10

Years Exp

EDA Advanced

7.0/10

Years Exp

Google Gemini AI Advanced

7.0/10

Years Exp

SQL Advanced

7.0/10

Years Exp

Google Colab Advanced

7.0/10

Years Exp

Jupyter Notebook Advanced

7.0/10

Years Exp

Git Advanced

7.0/10

Years Exp

AWS Advanced

7.0/10

Years Exp

Docker Advanced

7.0/10

Years Exp

Flask Advanced

7.0/10

Years Exp

LangChain Advanced

7.0/10

Years Exp

Power BI Intermediate

6.5/10

Years Exp

Tableau Intermediate

6.5/10

Years Exp

BeautifulSoup Intermediate

6.5/10

Years Exp

GitHub Actions Intermediate

6.5/10

Years Exp

MS Excel Intermediate

6.0/10

Years Exp

VS Code Pandas NumPy Matplotlib scikit-learn Postman Seaborn JavaScript NLTK MySql CSS HTML

Work Experience

AI/ML Engineer

Credit Card Fraud Detection

Present - Present

Trained a fraud detection system on 284,807 real transactions (Kaggle ULB dataset) with extreme class imbalance (0.17% fraud rate); applied SMOTE for balancing and engineered 9 features including cyclic encoding and z-score anomaly signals. Deployed a 3-model calibrated ensemble (isotonic calibration + recall-optimized thresholds) to maximize fraud detection while minimizing false positives. Built a Flask REST API with real-time prediction, risk classification (LOW / MEDIUM / HIGH), and contextual fraud reasoning.

AI/ML Engineer

Augmented Reality Navigation for Visually Impaired

Present - Present

Built an assistive AI navigation system to help visually impaired users identify obstacles in real-time. Used YOLOv8-based object detection to recognize surrounding objects such as vehicles, pedestrians, and barriers. Implemented depth estimation to calculate distance between user and detected objects. Built multilingual OCR module to extract environmental text like signboards and labels. Designed real-time voice feedback system to guide user navigation safely.

AI/ML Engineer

Virtual Try-On

Present - Present

Built an AI-driven fashion platform integrating web scraping, NLP-based product matching, and virtual try-on capabilities. Developed data extraction pipeline to collect and compare product data across multiple e-commerce platforms. Applied Natural Language Processing for similarity matching and intelligent price comparison. Implemented generative AI-based virtual try-on system for realistic product visualization. Integrated AI-powered recommendation support for enhanced decision-making.

AI/ML Engineer

RAG Pipeline Implementation

Present - Present

Built an end-to-end Retrieval-Augmented Generation (RAG) pipeline ingesting PDF and text documents with automated extraction, recursive chunking, and vector embedding generation for semantic search. Integrated Groq API for LLM inference with a vector retrieval layer, enabling context-aware question answering and reducing hallucinations by grounding responses in source documents. Implemented chunking strategies, sentence transformer embeddings, and similarity-based retrieval for accurate context extraction across large knowledge bases.