Khimraj Suthar
Email: khimrajsuthar@gmail.com
Website | LinkedIn | GitHub | Google Scholar | Resume
Summary
GenAI Engineer with 5+ years taking AI from prototype to production adoption — shipping multi-agent systems, RAG pipelines, fine-tuned models, and evaluation frameworks that employees and customers actively use at scale. Delivered solutions processing 4M+ records at 100K/day throughput across OpenAI, Claude, Gemini, and Llama ecosystems, with measurable impact on enterprise productivity and business outcomes.
Experience
GenAI Engineer
Bengaluru, India
Docusign
December 2024 - Present
- Designed a multi-stage LLM classification pipeline with model routing, confidence scoring, human-review workflows, and offline evaluation to track accuracy, processing 4M+ records at 100K/day throughput.
- Built shared GenAI components including LLM routing with cost and latency optimization, MCP services, agent authorization, RAG pipelines, and observability integrations that accelerated development across internal AI applications.
- Led development and rollout of Glean-based productivity agents, including automated daily action items, adopted by 50%+ of employees.
- Developed a Text-to-SQL pipeline using LangChain and vector databases, achieving 94% query accuracy and reducing reporting turnaround by 80%.
- Architected personalized email generation agents using CrewAI and LangChain, delivering 10K+ tailored emails and increasing click-through rates by 14%.
Software Engineer, LLM
Remote
Webb.ai
December 2023 – November 2024
- Architected multi-agent RAG systems for automated DevOps/SRE troubleshooting in cloud environments, handling 100 parallel requests per second in production.
- Fine-tuned LLMs using LoRA on tool-call datasets, built evaluation pipelines to benchmark function-call accuracy on held-out benchmarks, improving it from 78% to 89%.
- Drove a 60% reduction in DevOps/SRE debugging time and 50% cost savings through LLM-driven root cause analysis and incident resolution workflows.
Software Engineer, Machine Learning
Bengaluru, India
Ivy Homes
June 2021 - November 2023
- Led development of OCR, Entity Extraction, Document Classification, Text Clustering, and Pricing/Valuation models, reducing property valuation error from 15% to 8% MdAPE.
- Built an LLM-powered real estate chatbot automating 90% of off-hours user queries and implemented data pipelines processing 3M+ records with a self-hosted VPN security layer.
Machine Learning Intern
Remote
Skillbee
April 2021 – June 2021
- Created solutions for Job Recommendation and Relevance challenges.
- Worked on Deep Learning-based document classification which improved the CV relevance on the platform from 65% to 93%.
Software Development Engineer Intern
Remote
Amazon
May 2020 – July 2020
- Designed, developed, and deployed a responsive lender runbook website for Amazon Pay Later Service on AWS, reducing standard partner materials’ preparation and exchange time by 80%.
- Won a 3-day hackathon with 30+ participants, and secured a full-time job offer.
Publications
- Human Activity Recognition using Accelerometer and Gyroscope Data from Smartphones: Khimraj, P. K. Shukla, A. Vijayvargiya and R. Kumar, “Human Activity Recognition using Accelerometer and Gyroscope Data from Smartphones” 2020 International Conference on Emerging Trends in Communication, Control and Computing (ICONC3), India, 2020.
- Voting-based 1D CNN model for human lower limb activity recognition using sEMG signal: Vijayvargiya A, Khimraj, Kumar R, Dey N. Voting-based 1D CNN model for human lower limb activity recognition using sEMG signal. Phys Eng Sci Med. 2021 Nov 8.
Projects
LLM Multi-Agent based Stock Selection and Price Prediction Model
- Established a model for the India and US markets with daily, weekly, and monthly stock price predictions, including prediction confidence scores.
- Outperformed market index by 3% using fundamental, historical price, and news data.
Indoor Localization For 5g And Beyond(BTech Thesis Group Project)
- Proposed a novel MOIL deep learning approach for indoor localization using CSI data.
- The localization accuracy in the metric of mean distance error is within 0.01m in LOS and 0.02m in the NLOS scenario, which is 96% more precise than existing techniques.
Automated Tweets Report
- Developed a Python script for keyword-based tweet collection, sentiment analysis, and Excel report generation.
- Deployed the solution on AWS EC2 for daily scheduled executions with automatic email delivery using cron jobs.
Education
-
Malaviya National Institute of Technology (MNIT)
Jaipur, India
Bachelor of Technology - Computer Science and Engineering
CGPA - 7.70/10
August 2017 - June 2021
Skills
- LLM Systems: RAG, Multi-Agent Systems, Agent Orchestration, A2A, MCP, Function Calling, Prompt & Context Engineering, Fine-tuning, Model Evaluation, LLM Routing, Guardrails, Observability
- Frameworks & Models: LangGraph, LangChain, CrewAI, ADK, Glean, Vertex AI, Claude Code · GPT, Claude, Gemini, Llama
- Infrastructure & Data: AWS, GCP, Azure, Kubernetes, Docker, Airflow, Arize, CI/CD · PostgreSQL, Redis, Qdrant, Vector DB
- Languages & ML: Python, C++, SQL · TensorFlow, Scikit-Learn, XGBoost, Deep Learning, NLP, Computer Vision
Coursework
- Undergraduate: Design and Analysis of Algorithms, Data Structures and Algorithms, Database Management System Operating System, Computer Networks, Digital Image Processing, Natural Language Processing, Object-Oriented Analysis and Design, Deep Learning, Advanced Topics in Databases, Concurrent and Parallel Programming, Software Engineering
- Additional: Machine Learning, Mathematics for ML, Deep Learning Specialization, TensorFlow in Practice Specialization, Machine Learning in Production, Generative AI with Large Language Models, Trading Basics, Trading Algorithms, Advanced Trading Algorithms, Creating a Portfolio, Machine Learning for Trading Specialization