Professional Experience

Dailoqa

AI/ML Engineer
Mar 2025 – Present | Noida, India
  • •Agentic RAG Systems: Built an advanced RAG system integrating vector and graph retrieval with automated evaluations (RAGAS) and prompt regression tests.
  • •Hybrid Search Optimization: Engineered a hybrid retrieval architecture combining embeddings, BM25, and reranking, improving answer accuracy by 40%.
  • •Infrastructure & Deployment: Shipped a Dockerized FastAPI backend with dynamic caching and autoscaling, maintaining sub-5-minute environment setups.
  • •Standardization: Built a Model Context Protocol (MCP) layer to standardize tool-calling and agent orchestration, reducing cross-team integration effort.
  • •Data Pipelines: Implemented a cloud-agnostic ingestion pipeline (AWS, Azure, local) processing 10K+ documents using semantic chunking and dynamic index routing.
  • •LLM Safety: Designed and implemented safety mechanisms such as prompt-injection defenses and role-based access control (RBAC) for LLM-driven systems.

    • CoverQ Technology Inc.

    AI Engineer
    Jan 2024 – Mar 2025
  • •Fine-Tuning & Relevance: Fine-tuned domain-specific LLMs to improve response consistency; boosted precision for complex queries by 40% and overall relevance by 35%.
  • •Ticket Intelligence: Built a large-scale system analysing 5M+ Jira/Salesforce tickets for automated topic clustering and CRM reporting.
  • •Agentic Workflows: Designed multi-agent workflows using LangGraph, increasing operational efficiency by 25%.
  • •High-Precision Retrieval: Optimized vector indexing and query decomposition, reducing retrieval latency by 45%–50% for multi-part queries.
  • •Robotics Integration: Led Elite Robots’ EC-series arm integration, managing a 100GB+ sensor data pipeline using InfluxDB and Python.

    • National Institute of Technology Trichy (NIT Trichy)

    Summer Intern
    May 2023 – Jul 2023Trichy, Remote
  • •Conducted 6+ months of network lateral movement research, identifying and mitigating threats, and publishing actionable insights.
  • •Implemented deception & anomaly-detection ML techniques, reducing false positives by 15% and improving incident response accuracy.
    • Education

    Oriental Institute of Science and Technology

    2020 – 2024 | Bhopal

    Bachelor of Technology (B.Tech) - Computer Science and Engineering - (Data Science) GPA-8.65

    Projects
  • •Built a production-grade microservice supporting 5+ VLM providers with fallback routing.
  • •Extracted documents to Markdown/​JSON/​HTML/​TXT with OCR, table parsing, and image description.
  • •Containerized with Docker and published to Docker Hub, achieving <3s cold start and production-ready deployment
  • •Delivered complete project lifecycle: requirements gathering → architecture → development → testing → documentation → Docker/​PyPI publishing → GitHub release (v1.0.0)
  • •Created an ML evaluation/​selection toolkit improving experimentation speed by 25%.
  • •Published on PyPI, achieving 1,000+ downloads with seamless pip install integration.
    • Skills
    Programming: Python, Java, SQL
    AI/ML: LLMs, LangGraph, RAG, Autogen, Agentic AI, HuggingFace, NLP, MCP.
    Tools: Git/GitHub/Gitlab, CI/CD, Docker, Linux, RAGAS
    Frameworks: FastAPI, Streamlit, Langchain, Docling, scikit-learn
    Infra & Tools: Microsoft Azure, Azure AI Foundry, Amazon AWS.
    Databases: Vector DBs (ChromaDB, Pinecone, Opensearch), Neo4j, Relational
    Certificates
    Machine Learning by Andrew Ng, Harvard's CS50 - Python, Neo4j Certified Professional, AWS Machine Learning Foundations, Generative AI with LLMs, Introduction to LangGraph, All Certificates and Proiles ->
    Open Source Contributions

    IBM - Docling

  • •Table Intelligence: Engineered a feature to add page_no to TableCell within the core data model, enabling high-fidelity cross-page table reconstruction for RAG pipelines.
  • •Performance Optimization: Improved system efficiency by implementing lazy loading of transformers in the HybridChunker, reducing initial memory overhead during document processing.
  • •Serialization Reliability: Fixed critical bugs in the MarkdownTableSerializer, specifically disabling numparse to preserve trailing zeros in financial/​scientific data tables.