Bhuman Pandita
AI Engineer building and deploying multi-agent LLM systems, RAG pipelines, and agentic AI workflows at scale. Published researcher and ASCE award recipient with hands-on experience across Azure OpenAI, Semantic Kernel, MCP servers, LangChain/LangGraph, FastAPI, and async Python systems.
Professional Experience
Machine Learning Engineer (Assistant Manager)
IndiGo (InterGlobe Aviation)
Promoted to Machine Learning Engineer (Assistant Manager) in July 2026.
- Promoted to Machine Learning Engineer (Assistant Manager) in Jul 2026.
IAL - Digital & Data Science Program
IndiGo (InterGlobe Aviation)
Built and deployed production AI systems for airline loyalty, fraud detection, edge AI, and MRO automation.
- Owned end-to-end design and production deployment of a multi-agent MCP server using Semantic Kernel, Azure OpenAI, FastAPI, and FastMCP to automate BluChip loyalty profile updates across internal databases, reducing TAT from 12 days to under 1 minute with PII compliance and OCR-based government ID verification.
- Integrated intelligent workflow orchestration to enable consistent and error-free data synchronization.
- Collaborated with business and IT stakeholders to standardize update logic and ensure compliance.
- Architected and delivered a hybrid AI Loyalty Fraud Detection system integrating MS Graph API, Azure Cosmos DB, three Capillary APIs, and Navitaire, with six detection modules including LLM-driven geographic feasibility reasoning and self-connecting flight protection.
- Prototyped an Edge AI IoT asset tracker using a custom CNN vision model on Raspberry Pi with Go and Python backends, presented to senior engineering leadership.
- Developing an NLP/LLM invoice categorization pipeline using embedding retrieval and LLM classification to auto-tag MRO Non-Routine Card tasks across heterogeneous vendor invoices.
AI Engineer Intern
Swostitech Solutions
Focused on production RAG systems and LLM optimization.
- Built and optimized a production RAG pipeline over 1,000+ enterprise documents, benchmarking GPT-4, LLaMA-3, and Mistral-7B across chunking strategies and embedding models.
- Reduced end-to-end RAG inference latency by 30% at p95 through embedding model quantization, flat-to-HNSW vector-store migration, and query-layer caching.
Data Analyst Intern
Rigi, Bangalore
Automated creator transaction reporting and improved analytics workflows.
- Automated end-to-end transaction reporting for 100+ creators using pandas, openpyxl, and cron, saving approximately 20 finance-team hours per week and surfacing payout discrepancies automatically.
- Optimized PostgreSQL queries using window functions, CTEs, and index tuning, cutting Metabase dashboard load time by 30%.
Summer Intern
Ministry of Education, Government of India
Applied computer vision and remote sensing techniques to high-resolution imagery.
- Applied panoptic and semantic segmentation techniques to high-resolution imagery.
- Researched and implemented methodologies for enhancing remote sensing data visualization.
Summer Intern
Wadia Institute of Himalayan Geology
Applied machine learning to historical flood data and warning-system development.
- Created a flood warning system using machine learning techniques in R.
- Incorporated monthly and annual rainfall indexes to analyse historical flood data.
Technical Arsenal
Featured Projects
RAG Red-Teaming Pipeline
A four-stage adversarial AI evaluation framework for systematically probing RAG applications across 13 attack categories.

CycleGAN Image Dehazing
A CycleGAN-based image dehazing model trained from scratch on unpaired hazy-to-clear image translation.

CNN Sound Classification
Benchmarked CNN architectures on 6,000+ audio samples across 13 classes, improving accuracy through Mel-spectrogram preprocessing.

SWOT Analysis Generator
A Gemini-powered Streamlit tool for dynamic strategic assessments and recommendations.

Wheat Rust Detection using AI
Compared deep learning architectures for wheat disease classification, achieving 98.33% validation accuracy.

Meeting Summariser App
A Streamlit meeting summarizer using Whisper transcription and GPT-generated minutes and action items.
RAG Red-Teaming Pipeline
A four-stage adversarial AI evaluation framework for systematically probing RAG applications across 13 attack categories.

CycleGAN Image Dehazing
A CycleGAN-based image dehazing model trained from scratch on unpaired hazy-to-clear image translation.

CNN Sound Classification
Benchmarked CNN architectures on 6,000+ audio samples across 13 classes, improving accuracy through Mel-spectrogram preprocessing.

SWOT Analysis Generator
A Gemini-powered Streamlit tool for dynamic strategic assessments and recommendations.

Wheat Rust Detection using AI
Compared deep learning architectures for wheat disease classification, achieving 98.33% validation accuracy.

Meeting Summariser App
A Streamlit meeting summarizer using Whisper transcription and GPT-generated minutes and action items.
RAG Red-Teaming Pipeline
A four-stage adversarial AI evaluation framework for systematically probing RAG applications across 13 attack categories.

CycleGAN Image Dehazing
A CycleGAN-based image dehazing model trained from scratch on unpaired hazy-to-clear image translation.

CNN Sound Classification
Benchmarked CNN architectures on 6,000+ audio samples across 13 classes, improving accuracy through Mel-spectrogram preprocessing.

SWOT Analysis Generator
A Gemini-powered Streamlit tool for dynamic strategic assessments and recommendations.

Wheat Rust Detection using AI
Compared deep learning architectures for wheat disease classification, achieving 98.33% validation accuracy.

Meeting Summariser App
A Streamlit meeting summarizer using Whisper transcription and GPT-generated minutes and action items.
Honors & Awards
First Prize, ASCE International Conference (CISSC-2025)
Received 1st Prize at the UG Event on Challenges and Innovations for Sustainable Smart Cities for research on AI-powered UAV crop disease diagnostics.
Amazon ML Challenge 2024
Placed 166th among 75,000+ registered participants.
Research Publications
Recommendations
Bhuman worked with me @ Rigi as a Data Analyst Intern and showed impressive attention to detail working alongside me on our newly launched Gap Up App. His eagerness to learn and ability to pickup new technical skills would make him a valuable asset to any organisation.