Ahona Sarkar

Machine Learning Engineer · AI Systems & MLOps

Kolkata, India

@ahona_sarkar

LinkedIn GitHub

0 followers

🎓 B.Tech in Computer Science and Engineering at Institute of Engineering and Management · Graduating 2026

About

Machine Learning Engineer with experience building and deploying end-to-end AI systems, including fraud detection and graph-based identity resolution. Proficient in Python, PyTorch, and MLOps tooling, with additional experience in LLM applications, full-stack development, and production data pipelines.

Education

2026

B.Tech in Computer Science and Engineering

Institute of Engineering and Management

Computer Science · Aug 2022 – Jun 2026

Computer Science · Aug 2022 – Jun 2026

Skills

<3 months

C SQL RAG Java Shap NumPy React MLOps Python Pandas Node.js PyTorch XGBoost LangChain Express.js JavaScript Matplotlib TensorFlow scikit-learn Random Forest Explainable AI Neural Networks Prompt Engineering graph neural networks Natural Language Processing

Tools / apps / platforms

<3 months

Git Linux Docker MongoDB Apache Kafka GitHub Actions Amazon Web Services AWS

Languages

Native / Bilingual

Bengali Speak

Fluent

English Speak

Conversational

French Speak

Basic

Spanish Speak

Projects

Fraud Detection System

Python, XGBoost, Random Forest, DNN, SHAP, Streamlit, Kafka

Engineered 34 domain-specific features on the IEEE-CIS dataset (590,000+ transactions) and trained a Random Forest / XGBoost / DNN ensemble, lifting fraud precision by 18%. Integrated SHAP explanations into a Streamlit dashboard, surfacing top-5 risk factors per flagged transaction and cutting analyst review time by approximately 30%. Deployed a Kafka streaming pipeline with MLOps hooks, enabling real-time inference at under 100ms latency on live transaction data.
Digital Footprint Profiling via Graph Neural Networks

Python, PyTorch Geometric, GraphSAGE, NumPy, Pandas

Trained a GraphSAGE identity-resolution model on graphs with 50,000+ nodes, achieving an 88% F1 score — 23% higher than the rule-based heuristic baseline. Applied neighbor sampling to cut GPU memory consumption by 40%, reducing per-epoch training time from 18 minutes to 6 minutes.
Alumni Management Portal

React, Node.js, Express, MongoDB, FastAPI, Docker, REST API

Built a full-stack MERN application supporting profile management, event coordination, and peer networking, scaling to 500+ active alumni users; shortlisted for Smart India Hackathon 2025. Designed an AI-powered alumni-student matching engine using OpenAI embeddings, Pinecone vector search, and cosine similarity, and automated deployment with Docker and GitHub Actions CI/CD.
IGNIS-L1: Explainable Solar Flare Forecasting

Designing a physics-informed pipeline fusing Aditya-L1's SoLEXS and HEL1OS instrument data with Neupert-effect features for explainable flare nowcasting; built a dashboard prototype and submitted a concept proposal for ISRO's Bharatiya Antariksh Hackathon 2026 (PS15).

Courses & certifications

Introduction to Machine Learning · IIT Madras, NPTEL · 2026
Data Structures and Algorithms in Java · IIT Kharagpur, NPTEL · 2025
IBM AI Engineer Professional Certificate · IBM SkillsBuild Program · 2025
Deep Learning Specialization · DeepLearning.AI

🏆 Achievements & awards

Top percentile in NPTEL certification courses

Scored in the top percentile across NPTEL certification courses by IIT Madras and IIT Kharagpur.
Reviewed 15+ peer-reviewed AI/ML papers

Reviewed and analyzed 15+ peer-reviewed AI/ML papers on Graph Neural Networks, NLP, and Explainable AI, synthesizing findings into structured research notes.

Explore the Broxer community

💬 Q&A Ask about roles, salaries and interviews — and help others. 👥 Circles Join discussions on skills, companies and industries. 📰 Feed Follow updates and insights from the community.