Wasim M Ansari

About Me

Results-driven AI Solutions Engineer with a unique background in law and data science. I specialize in developing privacy-first AI systems that respect data protection standards while delivering measurable business impact. My legal and tech background ensures AI solutions that are not just technically sound but also ethically compliant and GDPR-adherent.

I specialize in LLM fine-tuning, RAG implementations, and production-grade ML model deployment. Currently undergoing rigorous training at IIT Madras in advanced ML/AI techniques, with hands-on experience in AI automation, vector databases, and generative AI solutions. My expertise spans PyTorch, TensorFlow, and Hugging Face's ecosystem for building scalable AI applications.

Featured Projects

• 
  MediFlow: Post-Discharge Medical AI Assistant [Live Demo] [GitHub]

  Problem: Patients discharged from hospitals often struggle to manage their medical care at home, lacking personalized guidance and easy access to accurate medical information.
  Solution: Implemented a multi-agent AI chatbot using LangGraph that combines RAG-powered knowledge retrieval, and real-time web search to provide personalized medical guidance tailored to patient's history.
  Impact: Delivers 24/7 intelligent medical support with 95%+ test coverage, enabling post-discharge patients to receive instant answers to their health concerns .
  Key Features: Multi-Agent System (Receptionist + Clinical agents), RAG Retrieval with MCP server tool, Real-time Web Search, Full LangSmith Tracing, Modern Web UI, LLM powered Testing.
  Tech Stack: Python, FastAPI, LangChain, LangGraph, MCP, Docker, Google Cloud Run.
• 
  Nephrology RAG MCP Tool [MCP Server Link] [GitHub]

  Problem: Medical knowledge bases need seamless integration with AI tools like Claude Desktop or custom MCP clients while supporting multi-user access with enterprise-grade security and scalability.
  Solution: Built a MCP (Model Context Protocol) server with HTTP/SSE transport that exposes medical knowledge, supporting concurrent users with session management and comprehensive monitoring.
  Impact: Enables Claude Desktop users and custom clients to access specialized nephrology knowledge, horizontal scaling to handle multiple concurrent users, and full observability.
  Key Features: MCP Server with HTTP/SSE Transport, Multi-user Session Management, CORS Support, Horizontal Scaling, Production-Ready Deployment (Docker, Cloud Run).
  Tech Stack: Python, FastAPI, MCP Protocol, FAISS Vector Store, Mistral AI, Docker, Google Cloud Run.
• 
  PDF Excel Extractor: AI Document Structuring [Live Demo] [GitHub]

  Problem: Businesses struggle to extract and structure unstructured data from diverse PDF documents (invoices, contracts, reports, forms), leading to manual data entry, errors, and inefficiency.
  Solution: Developed an AI-powered document processing system that dynamically detects relevant fields using LLM analysis, performs batch data extraction in a single API call, and generates structured Excel files with quality assurance metrics.
  Impact: Reduced document processing time by 80% with zero hardcoded schemas, enabling flexible handling and automatically adapting to new data structures without code changes.
  Key Features: Dynamic Schema Detection, Batch Processing, PDF Type Detection, REST API with Job Polling, Quality Assurance (BLEU scores & coverage metrics), Interactive Web UI, Excel Export.
  Tech Stack: Python, FastAPI, Mistral AI, LangChain, openpyxl, PyPDF2, Docker, Google Cloud Run.
• 
  DocuDroid: AI Document Analysis Bot [Live Demo] [GitHub]

  Problem: Coders often struggle to sift through large library and API documentation when they only need a concise code example or usage pattern. Reading long PDFs or web pages slows down development and integration tasks.
  Solution: Implemented a temporary in-memory vector store with chunking for immediate retrieval, combined with RAG and LLM-generated answers to surface precise code snippets and usage guidance on demand.
  Impact: Accelerated developer workflows by surfacing targeted code examples and reducing time-to-first-working-integration for common use cases.
  Key Features: Intelligent Chat with LLM, PDF & Web Content Analysis, Chunking & In-Memory Vector Store, Real-time Retrieval, Context-Aware Code Snippets.
  Tech Stack: Python, FastAPI, LangChain, Mistral AI, InMemoryVectorStore, Modern HTML/CSS/JS, Docker, Google Cloud Run.
• 
  Flux Image Generator [Live Demo] [GitHub]

  Problem: Many users must log in to use image-generation tools and lack API-driven automation, making integration into content pipelines and CI workflows difficult.
  Solution: Built a smooth, production-ready image generation platform that exposes REST API endpoints alongside an intuitive UI, integrating industry-standard diffusion models for both interactive and automated use.
  Impact: Streamlined automation and integration for design and content workflows, enabling headless batch generation and faster time-to-content for teams and CI/CD pipelines.
  Key Features: Text-to-Image Generation, REST API for automation, Multiple Model Support, Batch & Scheduled Generation, Custom Image Settings, Fast Processing, Immediate Downloads.
  Tech Stack: Python, FastAPI, Azure Web Apps, Hugging Face Diffusers, PIL, Docker.
  
• 
  RudeBot - Your Sassy AI Companion [Live Demo] [GitHub]

  Problem: Originally built as a weekend fun project — many chatbots are overly polite and standard. Sometimes users want humour, sarcasm or a rude persona to spark entertaining interactions instead of a strictly formal assistant.
  Solution: Implemented a context-aware chatbot with streaming responses and a deliberately rude, humorous personality to create lively, engaging conversations. This remains a prototype intended for experimentation rather than production deployment.
  Impact: Served as a demonstrative prototype for conversational persona design and streaming interactions; great for demos and experimentation but not intended for production use.
  Key Features: Streaming responses, Persona-driven replies (rude & humorous), Persistent memory, Threaded conversations, Context awareness.
  Tech Stack: Python, Mistral AI, LangGraph, Streamlit, Docker.
  
• 
  FaceMask Detection with CNN & VGG16 [Live Demo] [GitHub]

  Problem: COVID-19 created a need for automated mask compliance monitoring in public spaces.
  Solution: Implemented a transfer learning approach using VGG16 architecture for real-time mask detection with three-class classification.
  Impact: Achieved 95% accuracy in real-time mask detection, enabling efficient monitoring of safety protocols.
  Key Features: Real-time Detection, Three-class Classification, Transfer Learning, Data Augmentation, Model Fine-tuning.
  Tech Stack: Python, PyTorch, VGG16, OpenCV, NumPy, Matplotlib.
• 
  Mogambo Voice AI Scheduler [Live Demo] [GitHub]

  Problem: Manual calendar management is tedious and time-consuming. Inspired by Alexa and Siri, I wanted to build a conversational agent to manage my calendar through natural voice commands.
  Solution: Implemented Google OAuth for secure access to personal calendar data and built a voice-driven assistant for scheduling and reminders. The app is currently in testing mode (not verified or published) and used for personal chores only.
  Impact: Demonstrates the feasibility of voice-based calendar management for personal productivity, but not yet ready for public or production use.
  Key Features: Voice & Text Input, Google Calendar Integration (OAuth), Intelligent Event Matching, Clarification Workflow, Privacy & Security, Modern UI.
  Tech Stack: FastAPI, LangGraph, LangChain, Google Calendar API, SpeechRecognition, pydub, gTTS, Bootstrap, FullCalendar.
• 
  Jurisol: AI-Powered Indian Legal Assistant [Live Demo] [Docker Hub] [GitHub]

  Problem: Lawyers need a single place for thorough research and precedent search, while clients often want quick, free legal advice and opinions on their matters.
  Solution: Implemented a RAG system that searches Vector Store, online government databases for case laws, providing instant, relevant results. No user data is collected-ensuring privacy and safety.
  Impact: Streamlined legal research and made free, privacy-respecting legal information accessible to both professionals and the public.
  Key Features: Instant Legal Insights, AI-Powered Search, Indian Law Focus, Government Database Scraping, User-Friendly Interface, Transparent Citations, No Data Collection.
  Tech Stack: Python, FastAPI, Streamlit, ChromaDB, MistralAI, LangChain, TavilySearch, PyPDF2, BeautifulSoup4, Docker.
• 
  AI Interview Prep Assistant [Live Demo] [GitHub]

  Problem: Interview candidates often struggle to find tailored, realistic practice questions and actionable feedback for their target roles.
  Solution: Developed an AI-powered assistant that generates custom mock interviews and detailed answers using Mistral AI, simulating real interview scenarios.
  Impact: Enabled 100+ users to boost their interview skills and confidence through targeted, interactive practice sessions.
  Key Features: Custom Question Generation, Detailed Answers, Progressive Difficulty, Multiple Topics, Instant Feedback, User-friendly Interface.
  Tech Stack: Streamlit, Python, Mistral AI, Custom CSS.
• 
  BikeValuePro: ML Price Predictor [Live Demo] [GitHub] [Docker Hub]

  Problem: Buyers and sellers in the used bike market face uncertainty and inefficiency due to unreliable price estimates and lack of data-driven tools.
  Solution: Built an ML-powered price predictor using advanced ensemble models and feature engineering to deliver accurate, fair market valuations.
  Impact: Delivered 92% price prediction accuracy, empowering users to make confident, data-backed decisions and negotiate better deals.
  Key Features: Accurate Price Prediction, AI-Driven Insights, User-Friendly Interface, Smart Feedback, Market Context Analysis.
  Tech Stack: Streamlit, Python, Scikit-learn, XGBoost, ColumnTransformer.
• 
  AI-Powered Assignment Solver [Live Demo] [GitHub]

  Problem: Built for an IIT Madras competition where the challenge was to solve 10 data analysis and coding problems in just 30 minutes.
  Solution: Developed an automated assignment solver that autonomously tackled all problems, earning an A+ grade.
  Impact: Demonstrated the power of autonomous AI agents in competitive coding and data analysis, achieving top marks under time pressure.
  Key Features: AI-generated code execution, Multi-format file processing, Self-debugging workflow, Secure sandbox environment.
  Tech Stack: Python, FastAPI, Mistral API, Pandas, pdfplumber, openpyxl, numpy, PIL, Docker.
• 
  Photo Journal App [Live Demo] [GitHub]

  Problem: Many users want a reliable, cloud-native platform to create and share photo journals, but existing solutions often lack scalability and seamless storage.
  Solution: Designed and demonstrated a serverless photo journal app on Google Cloud Run, integrating managed storage and database services for robust, hands-off operation.
  Impact: Achieved 99.9% uptime and effortless scaling, allowing users to upload, view, and share photo journals with zero maintenance hassle.
  Key Features: Image uploads, Multi-user support, Chronological feed, Cloud-native architecture.
  Tech Stack: Flask, Google Cloud SQL, Google Cloud Storage, Google Cloud Run, HTML + CSS.
• 
  Text Similarity Detector [Live Demo] [Docker Hub] [GitHub]

  Problem: Content platforms struggle with duplicate question detection and semantic similarity analysis.
  Solution: Developed an ML-powered text similarity system using Word2Vec and XGBoost for accurate duplicate detection.
  Impact: Achieved 88% accuracy in duplicate detection, significantly reducing content moderation time.
  Key Features: Semantic similarity analysis, Confidence scoring, History logging, Smart ML pipeline, Modern dark UI.
  Tech Stack: Python, Flask, XGBoost, Word2Vec, Docker, HTML + CSS.
• 
  System Threat Forecaster (Kaggle Top 5%) [Kaggle]

  Problem: Kaggle competition organized by IIT Madras to develop an ML model for threat detection.
  Solution: Experimented with multiple model-building iterations and designed an ensemble technique that ranked in the top 5% on the leaderboard.
  Impact: Demonstrated advanced ML experimentation and ensemble modeling, achieving high accuracy and recognition in a national competition.
  Key Features: Feature engineering, Ensemble modeling, Kaggle competition ranking.
  Tech Stack: Python, Random Forest, XGBoost, Pandas, Scikit-learn.

Professional Experience

• AI/ML Engineer & Solutions Developer — Freelance/Independent Projects (Dehradun, India)
  Sep 2024 – Present
  • Pursuing IIT Madras Diploma in Data Science while rigorously self-learning advanced topics, with a strong interest in Transformers Architecture, Reinforcement Learning, and Cloud Services.
  • Designed and deployed multiple end-to-end AI/ML applications, spanning predictive modeling, conversational AI, legal research assistants, and voice-enabled scheduling agents - showcasing skills in XGBoost, LLMs, RAG, LangChain/LangGraph, and vector databases.
  • Delivered production-grade deployments using Docker, FastAPI, Flask, Streamlit, and Google Cloud/Azure, emphasizing scalability, UI/UX, and security.
  • Achieved Top 5% Kaggle ranking in a cybersecurity detection challenge, highlighting strong applied ML skills and problem-solving mindset.
  • Actively building a diverse project portfolio to demonstrate adaptability, technical depth, and eagerness to grow into real-world AI/ML challenges.


• Retail Manager — AHAAN LIMITED (Kingston Upon Hull, UK)
  Dec 2022 – Sep 2023
  • Spearheaded operational excellence initiatives and compliance programs across multiple retail locations, resulting in 20% revenue growth through strategic sales planning, inventory optimization, and enhanced customer experience metrics.
  • Pioneered the adoption of data-driven decision making by implementing Statistical-based solutions for dynamic pricing strategies and demand forecasting, leading to 15% reduction in stockouts and improved profit margins.


• Import Export Entrepreneur — AHLAN EXPORTS (Dehradun, India)
  Sep 2021 – Nov 2022
  • Successfully managed end-to-end international B2B logistics operations and regulatory compliance across multiple countries, while developing automated digital marketing funnels that increased exposure by 40%.
  • Leveraged data analytics for supply chain optimization and market trend analysis, resulting in 25% reduction in logistics costs and improved delivery timelines.


• Real Estate Consultant — FAHZAM PROPERTIES (Dubai, UAE)
  Feb 2015 – Jan 2020
  • Managed diverse property portfolios worth AED 50M+ and facilitated complex legal clearances as a certified RERA Broker, consistently achieving 95% client satisfaction rates through personalized property consulting.
  • Developed comprehensive market analysis reports and property valuation models, leading to successful closures of high-value transactions and establishing strong client relationships across the UAE.


• Legal Associate — QUISLEX LEGAL SERVICES (Hyderabad, India)
  Sep 2013 – Mar 2014
  • Led document review teams for complex litigation support projects, managing over 10,000+ documents monthly while maintaining 99% accuracy in legal analysis and classification.
  • Implemented systematic data discovery protocols and standardized reporting procedures, reducing processing time by 30% and improving case management efficiency.

Professional Standards & Certifications

Technical Expertise