Open to opportunities

Prashant Powar

@prashantpowar

Message

I build AI-powered web applications, from scam detection systems to computer-vision and LLM-driven interview platforms.

India

Message

What I'm looking for

I'm looking to grow in AI/ML and backend development by building practical, production-minded applications using Python, APIs, LLMs, and machine learning.

I've built AI-powered applications including S.I.R.A., a scam intelligence and response system that combines anomaly detection, behavioral biometrics, and Gemini-based classification to identify scam attempts with 92.5% accuracy. I also developed multi-agent workflows that extract actionable intelligence and deliver explainable real-time fraud alerts through FastAPI.

I'm a Computer Science undergraduate building across Python, FastAPI, React.js, TensorFlow, and databases. My projects include a skin disease prediction app, an AI mock interview platform with secure JWT authentication and Gemini feedback, and a YouTube sentiment classifier that achieved 85%+ accuracy.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Developer

Current

S.I.R.A.

Jan 2026 - Present (8 months)

Developed an AI-driven cybersecurity system combining anomaly detection, behavioral biometrics, and LLM-based classification, achieving 92.5% accuracy in identifying scam attempts. Built a multi-agent system powered by Gemini, spaCy NER, and Regex to autonomously engage threat actors and extract actionable intelligence.

Python fastAPI scikit learn MongoDB SpaCy Regex Anomaly Detection Behavioral Biometrics Multi Agent Systems

Skin Disease Detection & Remedy Recommendation System

Skinalyze

Jul 2025 - Nov 2025 (4 months)

Built a computer-vision web app that predicts skin conditions from uploaded photos and suggests remedies, with an end-to-end deep learning pipeline in TensorFlow. Deployed a Flask REST API for real-time, multi-class skin disease prediction.

Python TensorFlow Flask Computer Vision Deep Learning data preprocessing REST APIs

YouTube Comment Sentiment Analysis

Personal Project

Aug 2024 - Dec 2024 (4 months)

Built a text classification pipeline to classify YouTube comments by sentiment, achieving 85%+ accuracy. Benchmarked Logistic Regression, Random Forest, and Naive Bayes, improving accuracy by 18% via F1-score and precision-recall based model selection.

Python scikit learn Pandas NumPy TF IDF Logistic Regression Random Forest Naïve Bayes

AI Smart Mock Interview Platform

Personal Project

Developed a full-stack platform simulating realistic job interviews with AI-generated feedback, built with FastAPI, SQLAlchemy, and PostgreSQL. Implemented secure JWT authentication and resume upload with RESTful session APIs, integrating the Gemini API for personalized feedback.