Neha Marathe

AI/ML Engineer with 2.5 years of experience building production-ready Generative AI, LLM, RAG, and multimodal AI systems. Proficient in Python, FastAPI, LangChain, OpenAI APIs, PostgreSQL, MySQL, vector databases, and asynchronous backend development. Experienced in designing scalable AI services, Retrieval-Augmented Generation (RAG) pipelines, embeddings-based search, REST APIs, observability dashboards, and Dockerized deployments for enterprise applications.

JLPT N4 Certified with valid Japanese language certification and moderate communication capability in Japanese. Passionate about building reliable, context-aware AI applications on structured and unstructured data while collaborating in global and cross-cultural environments.

Associate Consultant — AI Engineer at Emergys Solutions Pvt Ltd (2024-01 – Present)

Formerly Ellicium Solutions

Associate Consultant — AI Engineer at Emergys Solutions Pvt Ltd (2024-01 – Present)

Project 1: Production-Ready Hybrid RAG System

• Designed a production-grade RAG pipeline combining semantic vector search (Chroma + OpenAI Embeddings) with BM25 keyword retrieval, fused via Reciprocal Rank Fusion (RRF) achieving ~15–25% higher retrieval precision/recall over single-method baselines.
• Implemented context-preserving chunking (800-char chunks / 200-char overlap), reducing information loss by ~30% versus naive splitting strategies.
• Developed asynchronous FastAPI endpoints for retrieval and generation workflows, improving response handling and supporting concurrent user requests.
• Built observability and monitoring dashboards using Grafana to track retrieval quality, response latency, hallucination metrics, and overall system health.
• Stack: LangChain (LCEL), Python, FastAPI, OpenAI GPT-4o, ChromaDB, PostgreSQL, BM25, Grafana, Docker

Associate Consultant — AI Engineer at Emergys Solutions Pvt Ltd (2024-01 – Present)

Project 2: E-Commerce Content Generation Platform

• Engineered a prompt-driven content generation pipeline using GPT-4o to produce product descriptions at scale, achieving 100% specification compliance with client brand guidelines.
• Built and deployed an analytics dashboard via Docker microservices enabling real-time content QA, reducing project delivery time by 30%.
• Developed multimodal AI workflows combining computer vision, OCR, and LLM-powered reasoning for intelligent engineering document understanding.

Associate Consultant — AI Engineer at Emergys Solutions Pvt Ltd (2024-01 – Present)

Project 3: Intelligent Diagram Analysis (Computer Vision)

• Trained YOLOv8 for object detection on complex P&ID engineering diagrams, reducing processing time by 40%; applied EasyOCR for text extraction with a 30% accuracy improvement.
• Implemented shape recognition and symbol classification using OpenCV, improving overall detection accuracy by 25%.
• Stack: YOLOv8, OpenCV, EasyOCR, Python, Hugging Face

BE in Computer Engineering in Computer Engineering – Savitribai Phule Pune University (2020 – 2024)