Mahendra O. looks like a good fit?

We can organize an interview with Aldin or any of our 25,000 available candidates within 48 hours. How would you like to proceed?

Schedule Interview Now

Mahendra O. - AI Engineer, Prompt Engineering, NLP

My name is Mahendra O. and I have over 3 years years of experience in the tech industry. I specialize in the following technologies: Data Science, Machine Learning, Deep Learning, Generative AI, Python, etc.. I hold a degree in Bachelor of Technology (BTech). Some of the notable projects I’ve worked on include: Llama Index Azure AI Search Advanced accurate Production level RAG, LangGraph Multi AI Agent Routing between (RAG + SQL + Normal Q&A), Accurate Advanced Agentic RAG, Fine Tuning LLM, Cost effective Earnings Call Transcript LLM Summarizer, etc.. I am based in Tadikonda, India. I've successfully completed 14 projects while developing at Softaims.

I value a collaborative environment where shared knowledge leads to superior outcomes. I actively mentor junior team members, conduct thorough quality reviews, and champion engineering best practices across the team. I believe that the quality of the final product is a direct reflection of the team's cohesion and skill.

My experience at Softaims has refined my ability to effectively communicate complex technical concepts to non-technical stakeholders, ensuring project alignment from the outset. I am a strong believer in transparent processes and iterative delivery.

My main objective is to foster a culture of quality and accountability. I am motivated to contribute my expertise to projects that require not just technical skill, but also strong organizational and leadership abilities to succeed.

Main technologies

  • AI Engineer

    3 years

  • lovable AI

    1 Year

  • AI Development

    2 Years

  • langchain

    2 Years

Additional skills

Direct hire

Potentially possible

Ready to get matched with vetted developers fast?

Let’s get started today!

Hire undefined

Experience Highlights

Llama Index Azure AI Search Advanced accurate Production level RAG

This advanced production RAG system uses Docling for accurate unstructured content extraction + chunking and used Azure OpenAI and text-embedding model. It retrieves documents from Azure AI Search with hybrid HNSW vector search, filtering by folder (relative_path) and metadata (file_name). A ChatEngine answers queries using only retrieved context, supported by Memory for long sessions. The system auto-detects relevant folders, dynamically constructs retrievers with hybrid search + reranking, logs, and filters sources by score threshold for accurate, clickable sources respect page numbers .

LangGraph Multi AI Agent Routing between (RAG + SQL + Normal Q&A)

i have built an intelligent agent system using LangGraph’s rag_sql_normal_Q&A_router to route natural language queries dynamically between a Retrieval-Augmented Generation (RAG) Agent , SQL Agent and normal Q&A Handler execution. This hybrid system interprets questions, identifies if they require database queries or pdf document retrieval, and returns rich, context-aware answers. Integrated LangChain, SQL databases, and vector stores to provide accurate, explainable results. Significantly improved data access, semantic understanding, and reduced hallucination in answer

Accurate Advanced Agentic RAG

Designed and deployed an intelligent document-based chatbot powered by Azure AI Search and GPT-4o. The system enables users to ask context-specific questions over selected PDF files. I integrated hybrid search (dense vector + keyword) and a semantic reranker for high-precision retrieval. The app leverages vectorized queries using OpenAI embeddings, filters content via parent_id, and delivers grounded answers using a RAG (Retrieval-Augmented Generation) approach and the UI was built using Streamlit for fast prototyping and interaction.

Fine Tuning LLM

Project Goals: Fine-tune the LLaMA model for instruction-based conversational AI using the Guanaco dataset, incorporating advanced techniques like QLoRA for memory efficiency and supervised fine-tuning to optimize performance. My Solution: Implemented LoRA for parameter-efficient fine-tuning, quantized the model to 4-bit precision for reduced GPU memory usage, and customized training configurations using `BitsAndBytesConfig` and cosine learning rate schedules. Successfully handled a structured dataset of instruction-response pairs to align the model for enhanced conversational tasks.

Cost effective Earnings Call Transcript LLM Summarizer

I developed an LLM-based summarization app Earnings call transcripts into concise summaries, highlighting financial metrics and strategic insights. 1) Data Preprocessing: - Scraped transcripts from sources like Motley Fool using BeautifulSoup. 2) Summarization Pipeline: - Utilized Azure OpenAI's GPT-3.5-turbo with token-based chunking. - Applied "stuffing" for chunk summaries and "refining" for cohesion. 3) Interface & Deployment: - Built a Streamlit app with cost tracking and URL-based inputs. 4) Cost Optimization: - Monitored token usage for budget-friendly operations.

Education

  • Lovely Professional University

    Bachelor of Technology (BTech) in Computer science Specialized in data science

    2020-01-01-2024-01-01

Languages

  • English
  • Hindi
  • Telugu

Personal Accounts