Headline

Data Analytics | Machine Learning | Python • SQL • PostgreSQL • Power BI

About

Computer Science graduate (2026) with experience in data analytics and machine learning — Python, SQL, Power BI. Built a two-layer fraud-detection system on 50,000+ e-commerce reviews — a fine-tuned DistilBERT classifier (87% accuracy) combined with a reviewer graph that independently catches ~35% of fake reviewers through timing and network patterns alone. Designed APGuard, a rule-based compliance system for Accounts Payable — 8 validation rules covering duplicate payments, contract price violations, and three-way match failures, tested against a synthetic 50,000-invoice dataset with pytest coverage. Also worked on satellite-based urban expansion mapping at MRSAC, using Google Earth Engine to detect land-use change across 15+ Maharashtra districts. Open to Data Analyst, BI Analyst, or Junior Data Scientist roles.

Location

Nagpur

AI stack

antigravity

claude

kimi

ollama

ChatGPT

Qwen

Work experience

Research Data Analyst at Maharashtra Remote Sensing Application Centre - MRSAC, Dec 2025 - Mar 2026 · 4 mos, Nagpur, Processed multi-temporal satellite imagery across 15+ Maharashtra districts using Google Earth Engine to map and measure urban expansion over a 5-year window Built an early warning module flagging encroachment hotspots via change-detection algorithms on LULC datasets Piped live AQI, thermal anomaly, and UV data into an environmental analytics backend, adding an OpenAI-powered query layer for plain-English access

Data Analyst Intern at Dealintra Infotech Pvt. Ltd., May 2025 - Jun 2025 · 2 mos, Nagpur, Built a two-stage review authenticity system combining a fine-tuned DistilBERT classifier with graph-based behavioural analysis to identify suspicious reviews and coordinated manipulation across Amazon review data. Engineered reviewer interaction graphs using timing, text similarity, and product overlap; Louvain community detection surfaced 132 suspicious clusters, independently identifying ~35% of fake reviewers through behavioural signals. Identified and corrected a train/test data leak (>95% overlap), rebuilt the evaluation pipeline with leak-safe splits, and added SHAP explanations to improve model transparency.

Education

Bachelor of Technology (B.Tech), Computer Science and Engineering, SB Jain Institute of Technology Management and Research, Yerla

Public links

github: https://github.com/smitkohale

smitportfolio-six.vercel.app: https://smitportfolio-six.vercel.app/