Ujala Maurya
Resume

Senior Machine Learning Engineer · Quantiphi · Mumbai, India

Building production-ready AI agents and intelligent systems.

I work on agentic AI, LLM-powered applications, enterprise RAG and guardrails. I build systems that connect AI capabilities with real production workflows.

View Projects Download Resume

Google ADK · Gemini · Vertex AI · LangChain · LangGraph · FastAPI · VespaDB · Python · GCP · AWS

40%
improvement in resolution efficiencyCare AI Agents
50%
reduction in AI response latencyCare AI Agents
40+
languages supported in productionCare AI Agents
40%
reduction in LLM hallucinationsAstra AI Platform

Featured projects

Two production systems. Outcomes are as reported on my resume.

Care AI Agents

Quantiphi · Nov 2025 – Present

Production AI agents that handle user queries autonomously across 40+ languages, with a governance layer that keeps agent behaviour controlled in live traffic.

  • 40% improvement in resolution efficiency
  • 30% improvement in response accuracy (Guard Agent)
  • 50% reduction in AI response latency
Technical details
What I built
  • Multi-agent orchestration. Contributed to the pipeline of issue-navigation, core-response and action agents built with Google ADK, running on Gemini and Vertex AI on GCP.
  • Guard Agent. Filters irrelevant queries and governs agent actions, enforcing controlled action execution.
  • Guardrails framework. Real-time detection of policy violations and sensitive topics, with configurable escalate, reject and resolve fallbacks across live and preview modes.
  • Multilingual support. Led end-to-end handling across 40+ languages: language detection, conversation handling and graceful resolution of language mismatches.
  • Latency. Introduced Pub/Sub-based asynchronous processing and deployed services on Cloud Run.
  • Retrieval and prompts. Owned LLM prompt design and knowledge-base / FAQ retrieval through the internal middleware layer (Setu).
  • Observability. Monitored live production data in Grafana to speed up debugging across AI services.
  • Google ADK
  • Gemini
  • Vertex AI
  • GCP
  • Cloud Run
  • Pub/Sub
  • FastAPI
  • Grafana

Astra AI Platform — Agentic RAG Server

Seclore Technologies · Jun 2022 – Nov 2025

An internal Agentic RAG chatbot for enterprise teams, giving fast, contextual answers from private datasets and taking routine deployment and production questions off engineering teams.

  • 65% improvement in response accuracy from hybrid search and re-ranking
  • 50% increase in platform adoption across teams
  • 40% reduction in LLM hallucinations
  • 60% less dependency on engineering teams for deployment and production queries
Technical details
What I built
  • Agentic RAG pipeline. Architected and implemented retrieval across multiple private document sources per team, with distributed query handling, using LangChain, FastAPI and VespaDB.
  • Retrieval quality. Implemented hybrid search and re-ranking.
  • Hallucination control. Prompt engineering and advanced data filtering.
  • Platform. FastAPI REST APIs and a Next.js frontend, with tool calling and Hugging Face models, on AWS.
  • Automation. Automated resolution of deployment and production queries.
  • LangChain
  • FastAPI
  • VespaDB
  • AWS
  • Hugging Face
  • Tool calling
  • Next.js

Experience

  1. Nov 2025 – Present

    Senior Machine Learning Engineer

    Quantiphi Analytics Solutions Pvt. Ltd. · Mumbai

    • Engineer and deploy AI agents on GCP with Google ADK, Vertex AI and Gemini for autonomous decision-making and multilingual query handling.
    • Built a Guard Agent and a real-time guardrails framework with configurable escalate, reject and resolve fallbacks.
    • Cut AI response latency by 50% with Pub/Sub-based asynchronous processing on Cloud Run.
    • Own LLM prompt design and knowledge-base / FAQ retrieval through the internal middleware layer; monitor live services in Grafana.
  2. Jun 2022 – Nov 2025

    Product Engineer (Generative AI)

    Seclore Technologies Pvt. Ltd. · Mumbai

    • Architected the Astra AI Platform's Agentic RAG pipeline over multiple private document sources with LangChain, FastAPI and VespaDB.
    • Improved response accuracy by 65% with hybrid search and re-ranking; reduced LLM hallucinations by 40%.
    • Built the REST APIs and Next.js frontend, increasing adoption by 50% across teams.
    • Automated deployment and production-query resolution, reducing dependency on engineering teams by 60%.

Technical expertise

Programming
Python, JavaScript, Java, SQL
AI and ML
Generative AI, LLMs, RAG, Transformers, NLP, Machine Learning, Hugging Face, Prompt Engineering
Agentic AI and application development
Google ADK, LangChain, LangGraph, FastAPI, Multi-agent systems, Tool calling, Pandas
Cloud and infrastructure
GCP, Vertex AI, Gemini, AWS, Pub/Sub, Docker, Jenkins, CI/CD, Linux Shell
Monitoring and engineering tools
Git, Grafana, SonarQube

About

I apply LLMs and AI agents to real operational problems. My work spans enterprise document retrieval, multi-agent systems, production APIs, AI safety and cloud deployment.

At Quantiphi I build and ship agentic systems on Google Cloud. Before that, at Seclore, I spent three years building generative AI products, from the first RAG prototypes to a platform used across teams.

B.E. Computer Science, St. Francis Institute of Technology, University of Mumbai (2018–2022).

Contact

Send me a message about a role, a project, or agentic AI and RAG systems. I'll reply by email.

Email directly Download Resume