Udeshya Raj

Data Scientist

Bengaluru, Karnataka, India0 mo experience

AI EnabledAI ML Practitioner

Key Highlights

Expert in AI-driven solutions and data science.
Proven track record in improving model efficiency.
Strong analytical mindset with real-time data processing.

Stackforce AI infers this person is a Data Scientist specializing in AI and Machine Learning solutions across various industries.

Contact

Skills

Core Skills

Machine LearningComputer VisionData ScienceNatural Language ProcessingProduct DevelopmentInternet Of ThingsBusiness Analysis

Other Skills

BERT embeddingsBERTweetBH1750LBeautifulSoupBusiness StrategyC (Programming Language)C++CatBoostClustering algorithmsCommunicationComputer ScienceData AnalysisData ModelsData VisualizationDeep Learning

Experience

0 mo

Total Experience

Average Tenure

Current Experience

Nobroker.com

Associate Data Scientist

Jul 2024 – Present · 1 yr 11 mos · Bengaluru, Karnataka, India · On-site

Built a classifier-based recommender system using CatBoost and BERT embeddings, improving click-to-contact ratio by 25–30% in Tier-1 cities. Deployed an end-to-end pipeline serving 1.2M+ users daily, including data mining, Prefect DAGs optimized with multiprocessing and chunking, Elasticsearch storage, and GKE-based APIs
Replaced an object detection system with Flash 1.5, significantly improving recall from 0.15 to 0.85 while reducing computational costs by 40%, enhancing real-time processing and efficiency in computer vision applications
Trained EfficientNet & Inception-ResNet for Neural Image Assessment, leveraging Generative AI (GenAI) to rank and score property images, improving image quality assessment, visual search, and recommendation accuracy
Built a User Profiling framework, applying machine learning (ML), clustering algorithms, and behavioral analytics to deliver personalized content recommendations, optimize lead generation systems, and drive user engagement and conversion rates
Developing a Text-to-Speech (TTS) system using Vall-E 2 & XTTS, reducing training time and data requirements while improving speech synthesis quality, multilingual adaptation, and real-time inference efficiency
Deployed DeepSeek R1 671B Distill, optimizing large language model (LLM) inference pipelines, fine-tuning transformer architectures, and ensuring scalability for AI-driven applications, including natural language understanding (NLU) and conversational AI

CatBoostBERT embeddingsElasticsearchGKEFlash 1.5EfficientNet+7

Finsire

Data Science Intern

Jan 2024 – Jul 2024 · 6 mos · Chennai, Tamil Nadu, India · Hybrid

Constructed Loan-to-Value (LTV) predictive models for asset valuation, leveraging advanced drawdown methodologies and machine learning algorithms like Random Forest and CatBoost
Attained an R² score of 0.96 through feature engineering, and hyperparameter tuning, and developed a data model for car resale value prediction
Implemented sentiment analysis using Generative AI (GenAI) models, creating a sentiment scoring system (0 to 10) to derive investment insights through natural language processing (NLP) techniques
Executed data wrangling and preprocessing on real estate datasets, and deployed RESTful APIs using the Tornado server for real-time property analytics across 800+ pin codes
Engineered ETL pipelines to perform automated web scraping, integrating multi-source data ingestion and storage on Google BigQuery

Loan-to-Value modelsRandom ForestCatBoostSentiment AnalysisRESTful APIsGoogle BigQuery+2

Saarthee

Data Science Intern

May 2023 – Jul 2023 · 2 mos · Gurugram, Haryana, India

Compiled and annotated dataset of 1000+ emails for training a custom T5 NER model, achieving 85%+ accuracy
Automated data extraction by creating an ETL pipeline using Selenium for web browser automation and BeautifulSoup for HTML parsing
Created APIs to retrieve info from the official Websites and implemented GCP (Google Cloud Platform) to validate findings

T5 NER modelSeleniumBeautifulSoupGCPNatural Language Processing

Singhal enterprises

Research And Development Intern

Jan 2023 – Apr 2023 · 3 mos · Kolkata, West Bengal, India

Developed a dynamic Python-based system that adeptly compares prices of different agricultural drone components and their suppliers
Crafted a Tender Identification System having >90% accuracy, enabling precise detection of potential government tenders
Identified the end customer needs, demand, and supply gap in the Agriculture segment by studying more than 100 drone startups in India

PythonMarket ResearchBusiness AnalysisProduct Development

Multimedia university

Research Intern

May 2022 – Jun 2022 · 1 mo · Malaysia

Assembled ESP32 microcontroller with INA219 current sensor and BH1750L light sensor to build a Low-cost solar power monitor
Analysed the impact of the shading effect on power output and reviewed the of parallel-series solution to minimize the power loss

ESP32INA219BH1750LInternet of Things

Iit kharagpur - vinod gupta school of management

Research Intern

Nov 2021 – Jul 2022 · 8 mos

Performed Sentiment Analysis using BERTweet with accuracy more than 85% after extracting and pre processing large volume of twitter data(> 100GB) from Twitter API.
Obtained narratives related to cryptocurrencies by applying the the short text topic modelling algorithm Gibbs Sampling Dirichlet Multinomial Mixture.
Analysed causal relation between crypto asset prices narratives through Convergence Cross Mapping bench marked against Granger Causality Test.

Sentiment AnalysisBERTweetGibbs SamplingNatural Language Processing

Rogan

Strategy Intern

Aug 2021 – Jan 2022 · 5 mos · Ahmedabad, Gujarat, India

Onboarded 80+ Businesses in Ahemdabad by implementing cross-promotion business strategy and collaborating with marts and schools
Researched the market dynamics and Forecasted annual reports of 3 years for pitching using the Top-Down method of forecasting

Market ResearchBusiness StrategyBusiness Analysis