Shlok Gopal Bhandari

About

Third-year Artificial Intelligence and Data Science undergraduate with hands-on experience building end-to-end ML and AI systems, including RAG pipelines, NLP applications, and financial-planning workflows. Skilled in Python, vector databases, and LLM integration, with current focus on AI agents and production-style pipelines.

Experience

• Data Science Intern
  NullClass EdTech Pvt Ltd
  Jan 2026 – Apr 2026

  Built intelligent chatbot systems using NLP, semantic search, and retrieval-based question-answering techniques. Implemented RAG-based knowledge retrieval pipelines using vector embeddings for context-aware, multi-turn chatbot responses. Applied sentiment analysis and multi-language capabilities; explored multi-modal AI (text and image) applications.

• Artificial Intelligence Intern
  Codec Technologies Pvt. Ltd.
  Jan 2026 – Feb 2026

  Developed Python-based NLP pipelines for text classification and data preprocessing; applied ML algorithms (classification and regression) and assisted in designing AI-based response systems.

Education

• 2027 PE
  Bachelor of Engineering in Artificial Intelligence and Data Science

  PES Modern College of Engineering

  Artificial Intelligence and Data Science · Aug 2023 – Aug 2027

  Artificial Intelligence and Data Science · Aug 2023 – Aug 2027

Skills

Tools / apps / platforms

Languages

Projects

• California Housing Price Prediction ↗
  Python, Scikit-learn, Pandas

  Built ML pipeline (imputation, scaling, encoding, stratified sampling, cross-validation) with Linear Regression, Decision Tree, and Random Forest; achieved 89% R2 on 20,000+ records.

• VidQuery AI ↗
  Python, Streamlit, Faster-Whisper, FFmpeg, Groq API, Ollama, RAG

  Built a full-stack RAG web app using Streamlit with drag-and-drop video upload, one-click pipeline automation, built-in media player, and chat interface with session-persistent history. Engineered end-to-end pipeline: video → audio → Faster-Whisper transcription (INT8) → BGE-M3 embeddings → cosine similarity retrieval → Groq LLM (LLaMA 3.3 70B) with Ollama fallback; reduced query retrieval time by ~40% and transcription time by ~60%. Added multilingual support to transcribe Hindi audio and translate to English via Faster-Whisper translate task; implemented skip-if-exists across all pipeline stages to avoid redundant reprocessing of already-indexed videos.

• WealthIQ ↗
  Python, Scikit-learn, Groq API

  Architected and orchestrated an 11-stage financial-planning pipeline coordinating deterministic engines and 2 LLM agents behind a single backend API, with a shared JSON contract enabling an 8-person team to build in parallel. Built a numeric-integrity verification layer that extracts every figure from LLM-generated explanations and cross-checks it against the source payload, rejecting hallucinated numbers and falling back to a safe template on failure. Designed the privacy-sanitization boundary stripping customer names, IDs, and raw transaction data before any LLM call, so AI agents only ever access pre-computed aggregate values. Integrated a Random Forest behavioral-risk classifier and Monte Carlo/stress-testing engines (10,000 simulated paths, 165 market-shock combinations per plan) to surface mismatches between stated and actual customer risk tolerance.

Courses & certifications

• AI Fluency for Students · Anthropic · 2026
• Artificial Intelligence Fundamentals · IBM SkillsBuild · 2026
• The Ultimate Job Ready Data Science Course · CodeWithHarry · 2025
• Python Training · Spoken Tutorial Project, IIT Bombay · 2024