Abhishek R.

Key Skills

Software

Don't disclose

Top Subject Matter

Nutrition recommendation and fish disease detection (LLM chatbots)

Cryptocurrency volatility prediction

Fraud detection automation using LLM prompting

Top Data Types

Text

Image

Top Task Types

Text Generation

Fine-tuning

Prompt + Response Writing (SFT)

Classification

Freelancer Overview

AI/ML Engineer | WebGen Technologies | Full-Time. Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Hugging Face, Don't disclose, and OpenAI API. Education includes Diploma in Data Science with Generative Artificial Intelligence, Physics Wallah Institute (2026) and Bachelor of Science in Physics (Honors), West Bengal State University (2025). AI-training focus includes data types such as Text and labeling workflows including Text Generation, Fine-tuning, and Prompt + Response Writing (SFT).

Labeling Experience

AI/ML Engineer | WebGen Technologies | Full-Time

TextText Generation

Developed and deployed LLM-powered nutrition and fish-disease chatbots, using generative AI workflows to produce domain responses from user prompts. Built application logic around prompt handling and agent behaviors that rely on structured text interactions for AI output quality. Worked with Llama 3, Hugging Face models, and Django/FastAPI backends to operationalize generative inference for production use cases. • Implemented prompt + response generation flows for chatbot interactions. • Integrated AI Agents logic with FastAPI and Docker deployment. • Used Hugging Face tooling to run and manage model components. • Delivered chatbot capabilities for health and aquaculture diagnosis-style use cases.

2026 - Present

AI/ML Engineer Intern | National Institute of Electronics & Information Technology (Jadavpur University)

Don't discloseTextFine-tuning

Developed cryptocurrency volatility prediction models using feature engineering and machine learning optimization over 20K+ records. Although the resume does not describe manual annotation, the work involved preparing structured training datasets and improving model performance for forecasting. Focused on transforming raw data into model-ready inputs and optimizing learning behavior to increase predictive accuracy. • Applied feature engineering to create predictive signals from records. • Trained/optimized ML models to reach reported accuracy. • Prepared and cleaned datasets for model ingestion. • Validated improvements via evaluation workflows.

2026 - 2026

Data Scientist Intern | Shamgar Software Solutions

TextPrompt + Response Writing (SFT)Classification

Built fraud detection systems using OpenAI API with prompt chaining and machine learning workflows for intelligent automation. The work centers on creating and using prompt-driven model behaviors to support fraud-related decisioning. Used LLM prompting strategies to structure multi-step reasoning and operational outputs. • Designed prompt chains to guide OpenAI API for fraud detection. • Combined LLM-driven workflows with ML automation pipelines. • Implemented end-to-end logic for intelligent fraud screening. • Tuned system behavior through iterative testing of prompt workflows.

2025 - 2026

Data Scientist Intern | AISECT

Text

Built a Myntra review scraping and analytics pipeline processing 50K+ reviews to deliver sentiment insights and business intelligence. The work involved transforming review text into analytics-ready representations to support sentiment evaluation. Used text analytics over large datasets to quantify sentiment signals for downstream decision-making. • Collected and processed 50K+ review text records. • Performed sentiment-driven analytics and derived insights. • Built a Python automation pipeline for ingestion and analysis. • Produced business intelligence outputs for stakeholders.

2025 - 2025

Data Scientist Intern | EvoAstra Ventures

Don't discloseTextClassification

Developed and deployed Visa fraud detection models over 500K+ records, improving accuracy by 22% and reducing manual preprocessing by 30%. This required building supervised learning pipelines for classification and preparing consistent training inputs from large datasets. Focused on model training, optimization, and dataset preparation to reduce annotation/preprocessing burden. • Engineered features and built classification models for fraud. • Optimized training workflows to improve predictive performance. • Reduced manual preprocessing effort through automation. • Validated improvements using accuracy and performance measurements.

2025 - 2025

Education

Work History