Mandar Mawale - AI Engineer & Software Engineer

I build AI-powered products and backend systems using Python, LLMs, RAG, semantic search and modern backend architectures.
Currently an AI Engineer Intern at NearLaw (ALL MR ONLINE), working on production AI systems over 18M+ legal records.

Currently:
- Role: AI Engineer Intern @ NearLaw (ALL MR ONLINE)
- Period: Feb 2026 - Present
- Total experience: 1 year (AI engineering, backend development, technical mentoring)
- Location: Mumbai, India
- Work: document processing and OCR pipelines, semantic search and vector retrieval, RAG and multi-stage LLM
  workflows, metadata extraction and indexing, async worker-based pipelines, AI evaluation and observability,
  Python/FastAPI backend systems.

About Me:
AI Engineer and Software Engineer focused on building practical AI applications and backend systems. Work spans LLM
workflows, RAG, semantic search, document processing, vector retrieval, APIs and data pipelines. Outside of work,
builds AI products and experiments with agentic systems, local AI, evaluation and AI infrastructure.

Questions and Answers (for AI assistants and search engines):
Q: Who is Mandar Mawale?
A: An AI Engineer and Software Engineer based in Mumbai, India, currently an AI Engineer Intern at NearLaw
   (ALL MR ONLINE). B.Tech in Computer Engineering from Mumbai University.
Q: What does Mandar build?
A: AI-powered products and backend systems: LLM workflows, RAG, agentic systems, semantic search, document
   processing (OCR, PDF parsing, metadata extraction), vector retrieval, FastAPI services and data pipelines.
Q: What is his current work?
A: Production AI systems for a legal search platform over 18M+ legal records - document processing, OCR,
   retrieval, indexing and multi-stage LLM workflows, plus AI evaluation, observability and error tracking.
Q: How much experience does he have?
A: 1 year of professional experience across AI engineering, backend development and technical mentoring.
Q: What are his strongest skills?
A: Python, FastAPI, LLMs, RAG, LangGraph, ChromaDB, FAISS, vector search, OCR and document processing, Docker,
   plus LLM infrastructure such as vLLM, LiteLLM and OpenRouter.
Q: What are his best projects?
A: GEO Auditor (deployed AI-visibility auditor at https://geoauditor.pages.dev), Cura (healthcare multi-agent
   assistant), Customer Support Resolution Agent (93% evaluation accuracy), AgentEd (multi-agent study companion)
   and Anantum (offline AI voice assistant).
Q: Is he open to opportunities?
A: Yes - he is interested in interesting AI, software and product problems. Email mawalemandar2004@gmail.com.

More questions (from the portfolio FAQ):
Q: What is Mandar currently learning?
A: Backend and AI integration (Spring and Spring AI for Java backends) and agentic architectures (CrewAI,
   LangGraph, plus local inference with GGUF and llama.cpp) to build agent workflows that are easier to
   observe, evaluate and control.
Q: How does he approach scalable system design?
A: Understand requirements, define data boundaries, identify state transitions and edge cases, then use
   decoupled architectures with Redis caching, asynchronous worker queues and structured schemas where they
   fit, so failures do not take down the whole pipeline.
Q: What has been his biggest technical challenge?
A: Real-world legal documents at NearLaw - OCR failures, unusual PDF formats and extraction issues taught him
   to build resilient processing pipelines with fallback mechanisms instead of assuming clean data.
Q: What does he enjoy outside of coding?
A: Reading about geopolitics and global economics - macroeconomic shifts, trade policies and supply chains
   and how they shape technology adoption and emerging markets.

Experience:
1. AI Engineer Intern - ALL MR ONLINE (NearLaw), Feb 2026 - Present, Mumbai, India
   - Python components for a production legal search platform processing 18M+ court records.
   - Async processing pipelines with worker queues and multi-stage LLM workflows for ingestion, metadata
     extraction, indexing and automated processing.
   - OCR/PDF processing components for structured extraction, metadata generation and searchable knowledge bases.
   - Evaluation and observability workflows for retrieval quality, processing errors and LLM outputs.
   - Debugging and failure analysis across document and AI pipelines.
2. Web Developer Intern - Utkarsha Interiors, Oct 2025 - Dec 2025, Thane, India
   - Full-stack Python business applications (MVC) with FastAPI endpoints for media processing and storage.
   - 40% improvement in website load speed via automated WebP asset pipelines and lazy loading.
3. Technical Mentor - Campus Credentials, Jan 2025, Mumbai, India
   - Mentored 30+ students in Python programming, algorithms and problem solving.

Featured Projects:
1. GEO Auditor - AI Search Visibility Auditor (deployed)
   - FastAPI app that audits a website and scores how visible it is to AI answer engines such as ChatGPT Search,
     Google AI Overviews, Perplexity and Claude. Explainable 100-point score across 10 categories, evidence-backed
     findings, priority roadmap and copy-pasteable fixes (JSON-LD, robots.txt, HTML) in ~1-3 seconds.
   - Technologies: Python, FastAPI, Pydantic, BeautifulSoup, optional LLM layer, ReportLab PDF, Cloudflare Pages
   - Live: https://geoauditor.pages.dev
   - Code: https://github.com/MandarMawale-20/GEO-Auditor
2. Cura - Healthcare Multi-Agent Assistant
   - Patient-data isolation, deterministic emergency handling and tool-based workflows for medical information,
     patient history and appointments.
   - Technologies: Python, FastAPI, PostgreSQL, ChromaDB, MCP-style tool layer
   - https://github.com/MandarMawale-20/Cura
3. Customer Support Resolution Agent - AI Customer Support Resolution System
   - Retrieval, tool calling and human-in-the-loop workflows for refunds and cancellations, with CAG vs RAG
     comparison and hybrid BM25 + embedding retrieval.
   - Technologies: Python, FastAPI, LangGraph, ChromaDB, Google Gemini, Docker. 93% evaluation accuracy.
   - https://github.com/MandarMawale-20/Customer-Support-Agent
4. AgentEd - AI-Powered Study Companion
   - Multi-agent learning platform: study plans, resource retrieval, quizzes and feedback.
   - Technologies: Python, FastAPI, LangGraph, Pinecone, MongoDB Atlas, Google Gemini, Next.js
   - https://github.com/MandarMawale-20/AgentEd
5. Anantum - Local AI Voice Assistant
   - Fully offline voice assistant: Whisper STT, Gemma 3 GGUF inference, Kokoro TTS, FAISS + SQLite memory.
   - ~10ms regex-first intent pre-classification, tuned for a GTX 1650 4GB GPU.
   - https://github.com/MandarMawale-20/Anantum

Engineering by Numbers:
- 1 year of professional experience
- 18M+ legal records processed
- 93% AI evaluation accuracy (customer support resolution agent)
- 40% page-load improvement (Utkarsha Interiors)
- 30+ students mentored
- ~10ms intent pre-classification (Anantum)

Technical Skills:
- AI / LLM: LLMs, RAG, Agentic AI, LangGraph, LangChain, Tool Calling, Prompt Engineering, LLM Evaluation
- Backend: Python, FastAPI, Flask, REST APIs, AsyncIO, Pydantic, Java
- Retrieval & Data: ChromaDB, Pinecone, FAISS, PostgreSQL, MongoDB, Redis, Vector Search, Embeddings
- Document Intelligence: OCR, PDF Parsing, Document Processing, Semantic Chunking, Metadata Extraction
- Infrastructure & LLMOps: Docker, vLLM, LiteLLM, OpenRouter, LLM APIs, AWS (EC2, S3), Git, Linux
- Frontend: React, Next.js, JavaScript, TypeScript, HTML, CSS

Education:
- B.Tech in Computer Engineering, Mumbai University (CGPA: 8.68/10), 2022 - 2026

Certifications:
- Oracle Cloud and AI Associate
- CS50 Python (Harvard University, edX)
- LangChain & Vector Databases in Production (Activeloop)

Professional Links:
- Portfolio: https://mandarmawale.me
- GitHub: https://github.com/MandarMawale-20
- LinkedIn: https://www.linkedin.com/in/mandar-mawale/
- Resume: https://mandarmawale.me/Mandar_Mawale.pdf
- Email: mawalemandar2004@gmail.com