×

Hello, I'm

Parimal Kulkarni

Data Scientist & Generative AI Engineer

Passionate about leveraging AI and ML technologies to solve complex problems. Specializing in Generative AI, RAG systems, and ML/DL applications.

View Projects Contact Me View Resume

Technical Skills

My toolkit for solving complex data problems and building intelligent applications.

Programming

  • Python
  • C/C++
  • SQL
  • MATLAB

Data Science & ML

  • pandas, NumPy, EDA
  • Feature Engineering
  • Predictive Modeling
  • Statistical Analysis
  • Supervised & Unsupervised Learning
  • scikit-learn

Deep Learning & NLP

  • Transformers, RNNs, LSTMs, CNNs
  • TensorFlow
  • Text Classification, NER
  • Word Embeddings
  • Sentiment Analysis
  • Text Processing

Generative AI

  • LLMs (GPT-4, LLaMA2/3, DeepSeek)
  • LangChain, LangGraph
  • RAG, FAISS, ChromaDB
  • Vector DBs, Hybrid Search
  • Fine-tuning (LoRA/QLoRA)
  • Prompt Engineering

Tools & Deployment

  • Streamlit, FastAPI
  • AWS, Hugging Face Hub
  • Groq, CrewAI
  • OCR (Tesseract, OpenCV)
  • Data Pipelines
  • Model Serving

Work Experience

My professional journey building AI solutions and data-driven products.

Apr 2025 - Jun 2025

PowerCred Technologies

Data Scientist – HITL Intern

  • Developed a GenAI Drive assistant leveraging LangChain, Python, LLaMA3-70B, and RAG+FAISS to perform fast and accurate semantic search across 1M+ documents in under 2 seconds.
  • Boosted search precision by 35% via vector+metadata retrieval (e.g., owner, modified date, size).
  • Designed SOC 2–compliant architecture on GCP using OAuth 2.0 and Cloud Functions to support 500+ real-time queries/day.
  • Automated document ingestion from PDFs/Office using PyMuPDF, Tika, PyPDF2, achieving 92% accuracy and 60% speedup.
  • Built a Streamlit UI with real-time bounding-box validation and seamless MLOps integration.
Feb 2025 - Mar 2025

Britannia Industries Limited

Data Science Intern (Generative AI)

  • Built a GenAI chatbot with GPT-4, LLaMA 2, RAG, and ChromaDB, cutting support resolution time by 30%.
  • Fine-tuned LLMs using domain embeddings and vector DBs, improving response accuracy by 25% with less than 1s latency.
  • Optimized GenAI pipelines using Data Structures, Algorithms, and Deep Learning, increasing engagement by 20%.
  • Performed sentiment analysis on 10,000+ reviews using spaCy and NLTK to guide product strategy.
Dec 2024 - Jan 2025

Orinson Technologies Pvt Ltd

Machine Learning Intern

  • Applied ML algorithms to analyze 200,000+ records, improving model accuracy by 15%.
  • Preprocessed data and fine-tuned models, leading to a 20% decrease in processing time.

Projects

Showcasing my technical expertise through innovative solutions and applications.

Advanced PDF Chat Assistant

Python, LangChain, Groq API, HuggingFace

  • Built a 94% accurate RAG pipeline with FAISS & LLM chaining, boosting relevance by 40% using optimized chunking.
  • Created a modular Streamlit UI with multi-backend LLMs (LLaMA3, DeepSeek) and 100+ page PDF processing in real-time.
  • Integrated source-cited conversational retrieval, enabling follow-up history tracking and structured agent responses.
View on GitHub

AI Research Assistant

Python, LangChain, Groq API, ArXiv, Wikipedia, DuckDuckGo

  • Developed multi-source GenAI agent querying ArXiv, Wikipedia, Web; selected optimal LLM (LLaMA3/DeepSeek) for each query.
  • Engineered a retrieval pipeline with concise 500-char answers; improved semantic quality and ensured factual grounding.
  • Implemented a robust fault-tolerant agent cutting LLM failures by 25%, boosting response speed by 60%.
View on GitHub

Medical Data Extraction OCR

Python, Tesseract OCR, OpenCV, NLP, FastAPI, Streamlit

  • Created an end-to-end OCR system for 1,500+ medical records/day with 92% accuracy and fast, structured JSON outputs.
  • Reduced parsing errors by 30% using NLP, DS/Algo, RegEx, ensuring reliable downstream data processing.
  • Cut manual labor by 70%, showcasing ownership, autonomy, and engineering mindset in delivery of solution.
View on GitHub

Aerospace Investment Rating Analysis

Python, Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn

  • Analyzed 200+ aerospace companies using SFR ratings and data pipelines to drive investment insights.
  • Trained ML models (Random Forest, Decision Tree) with 87% accuracy for financial risk prediction.
  • Built visual dashboards for exploratory data analysis, aiding strategic decision-making and analytics.
View on GitHub

Osteoporosis Risk Prediction

Python, Scikit-learn

  • Built ML models (Decision Tree, SVC, Random Forest, Logistic Regression) on 1,000+ medical records to predict osteoporosis risk.
  • Achieved 87% accuracy and identified top 5 contributing risk factors for early intervention insights.
View on GitHub

Hackathon Achievements

Competitive problem-solving experiences showcasing creativity and technical prowess.

HackRx 5.0 Finalist (Team Lead)

Top 22 Teams

Bajaj Finserv Health

  • Utilized Azure Cognitive Services to extract medical diagnoses from unstructured text data with 88% accuracy.
  • Implemented FuzzyWuzzy for matching extracted diagnoses with ICD-10 codes, achieving a 93% success rate.
  • Leveraged Azure OCR and Gemini 1.5 Flash, improving text processing efficiency by 40%.

Hackademia (Team Lead)

Top 3 Teams

BMC Software

  • Developed a conversational RAG system with PDF processing, FAISS vector storage, and session management.
  • Built a RAG Validation Framework with 12 evaluation metrics (Precision >0.8, Recall >0.7).
  • Integrated ChatOpenAI, Google Generative AI, Groq, and BERT for retrieval and ethical validation.
  • Evaluated responses using BERTScore, ROUGE, BLEU, METEOR, cosine similarity, and FactCC.
  • Designed an interactive Streamlit dashboard with Plotly visualizations for performance analysis.

Courses & Certifications

Continuous learning to stay at the forefront of data science and AI technologies.

Python for Machine Learning and Data Science Masterclass

Udemy

CS50's Python

Harvard University

Data Science and AI

IBM

Crash Course in Python

Google

Generative AI with LangChain and Hugging Face

Udemy

RAG-LLM Evaluation & Test Automation

Udemy

Get In Touch

Interested in working together? Let's connect and discuss how we can collaborate.

Email

parimal.kulkarni@sitpune.edu.in

Phone

+91 8087768244

LinkedIn

Connect on LinkedIn

Download Resume

Download PDF
// Remove background GIF rotation code