Credit-Card Fraud Detection using Autoencoders
Summary
Developed a robust credit-card fraud detection system utilizing autoencoders to identify anomalies in transaction data.
About
Highly motivated AI/ML and Data Science practitioner with 2 years of hands-on experience, specializing in developing scalable, intelligent systems leveraging LLMs, Reinforcement Learning, and Generative AI. Proven expertise in prompt engineering, LangChain, Agentic AI, and RAG pipelines, with a strong drive to deliver high-impact ML solutions that address real-world challenges and drive significant efficiency gains.
Work
Veersa Technologies
|Data Science Intern
Noida, UP, India
→
Summary
As a Data Science Intern, Aryan developed and deployed advanced AI/ML solutions, focusing on LLM-based document intelligence, multi-agent systems, and NLP pipelines to drive significant operational efficiencies.
Highlights
Developed vfileManager, an LLM-based document intelligence system leveraging LangChain and FAISS, automating file clustering and semantic QnA to reduce document retrieval time by 60%.
Architected a multi-agent Agentic AI Assistant for medical recommendations, interaction checks, dosage validation, and automated report dispatch, cutting manual doctor workload by 50%.
Deployed Profanity Pulse, an LSTM classifier for 7-category text toxicity scoring, achieving 91% accuracy in content moderation.
Implemented vChat, fine-tuned GPT-3 with an internal RAG pipeline, increasing response relevance by 35% for conversational AI.
Led the Post-call Summary and Classification NLP pipeline for client analytics, cutting manual summary time by 65%.
Automated the EDC Platform for HealthCat by streamlining Databricks DAB orchestration, reducing effort by 70%.
Enhanced vExtract, a medical form parser using LLM + FastAPI, enabling 80% automation in data extraction.
Designed a Capacity Planner, a rolling-window prediction engine using XGBoost and SHAP, improving accuracy by 22%.
AI-Shala
|AI/ML Intern
New Delhi, Delhi, India
→
Summary
As an AI/ML Intern, Aryan trained and engineered advanced AI/ML models, including Reinforcement Learning agents, Seq2Seq translation models, and LLM guardrails, to solve complex challenges.
Highlights
Trained Reinforcement Learning (RL) agents (Q-Learning, PPO, SAC, DDPG) using d3rlpy, developing a Swing Trading RL Bot.
Constructed LSTM-based Seq2Seq and Transformer models for English-Hindi translation, achieving high accuracy in language processing.
Simulated a custom RL environment for Systematic Investment Planning (SIP) and explored TimeGPT for time series analysis.
Engineered LLM Guardrails for hallucination, toxicity, and code clone detection using Tree of Thoughts, enhancing model reliability.
Summer Training/Internship
|AI/ML Intern
New Delhi, Delhi, India
→
Summary
During this AI/ML Internship, Aryan applied machine learning techniques to design a fraud detection system and implement computer vision models, while performing foundational data analysis.
Highlights
Designed an Audit Fraud Detector for detecting fraudulent transactions, achieving 94.5% accuracy and a 0.96 AUC score.
Implemented CNNs, RNNs, GANs, and autoencoders for vision and anomaly detection tasks.
Performed dimensionality reduction, data cleaning, and exploratory data analysis on real-world datasets to prepare for advanced modeling.
Coincent
|Data Science Intern
New Delhi, Delhi, India
→
Summary
As a Data Science Intern, Aryan gained practical experience in core data science methodologies, including data cleaning, feature engineering, ML modeling, and model evaluation techniques.
Highlights
Applied comprehensive data cleaning, feature engineering, machine learning modeling, and model evaluation techniques to real-world datasets, enhancing predictive accuracy.
Education
Maharaja Surajmal Institute of Technology
Delhi, Delhi, India
→
Bachelor of Technology
Computer Science and Engineering
Languages
English
Hindi
Certificates
Azure Data Fundamentals
Issued By
Microsoft
Data Analytics
Issued By
IBM SkillsBuild
Microsoft Azure DP-900 certification in Data Fundamentals
Issued By
Microsoft
Skills
Programming Languages
Python, SQL, C, C++, JavaScript.
AI/ML Frameworks & Tools
TensorFlow, PyTorch, NumPy, Pandas, d3rlpy, LangChain, FastAPI, Flask, MLflow, Azure ML, XGBoost, SHAP, FAISS, Tree of Thoughts.
Data Analysis & Visualization
Power BI, Tableau, Microsoft Excel, Data Cleaning, Feature Engineering, Exploratory Data Analysis.
Machine Learning & AI Concepts
LLMs, Reinforcement Learning (RL), Generative AI, Databricks, ML Pipeline Automation, Prompt Engineering, Agentic AI, RAG Pipelines, NLP, Computer Vision, Anomaly Detection, Time Series Analysis, Fraud Detection, LSTM, CNNs, RNNs, GANs, Autoencoders, Seq2Seq, Transformers.
Cloud & DevOps
Git, Azure Data Fundamentals, Databricks DAB Orchestration.
Projects
Disease Prediction from Call Recordings between Patient and Medical Staff
Summary
Designed and implemented a system to predict diseases by analyzing call recordings between patients and medical staff, leveraging natural language processing techniques.