Skip to content
Khushal Jethava

About

Khushal Jethava

Machine Learning Engineer · GenAI, Voice AI & Multi-Agent Systems

At Drona's test, every student saw the tree, the branches, the leaves. Arjuna saw only the bird's eye. That's how I like to build AI: find the one thing that matters, and hit it.

  • Generative AI
  • Chatbot development
  • Natural Language Processing
  • Computer vision
  • Multi-agent orchestration
  • Real-time voice AI
  • RAG pipelines

Ahmedabad, Gujarat, India

Read the story

Chapter I · शंखनाद

Every journey starts with a call.

I'm a Machine Learning Engineer with a Bachelor's in Computer Applications from GLS University. Project planning, programming and hands-on technical projects there laid the foundation for my career in AI and machine learning.

Four-plus years on, I design and deploy production-grade Generative AI: LLM-powered applications, multi-agent orchestration, RAG pipelines, real-time voice AI and low-latency AI infrastructure, owning the full lifecycle from research and modeling to deployment, monitoring and optimization, and working closely with product, engineering and business teams.

Some of the work I'm proudest of:

  • A production multi-agent orchestration platform on LangGraph, with deterministic execution and strict phase isolation.
  • Real-time voice AI with sub-800ms end-to-end latency, handling 1K+ daily calls at 99.5% uptime.
  • Image segmentation for room transformation using ControlNet.
  • Fine-tuning segmentation models for optimized accuracy.
  • Building backend APIs to support complex AI functionality.
  • Creating a Document AI product on fine-tuned transformers.
  • Designing a barcode detection system with YOLO and OpenCV.
  • Building a car damage detection framework with Detectron2 and CNNs.
  • Leading generative chatbot projects with LSTM sentiment analysis and BART dialogue summarization.

Chapter II · कर्मभूमि

The field of work.

  1. Senior Machine Learning Engineer · Bytes Technolab

    April 2026 – Present · Ahmedabad, Gujarat, India · bytestechnolab.com

    Building computer vision, Generative AI, AI automation and multi-agent workflow solutions.

  2. Machine Learning Engineer · Codiste Private Limited

    April 2024 – March 2026 · Ahmedabad, Gujarat, India

    • Architected a production-grade multi-agent orchestration platform using LangGraph, with deterministic execution and strict phase isolation.
    • Designed orchestrator-driven agent routing across onboarding, planning, execution and tool-invocation stages for complex, multi-step workflows.
    • Implemented persistent shared state and unified confirmation logic, reducing redundant user interactions by 35% and lowering API costs.
    • Built real-time AI voice pipelines integrating STT, LLM reasoning and TTS with sub-800ms end-to-end latency.
    • Integrated OpenAI, ElevenLabs and telephony (SIP, Twilio, PSTN) to handle 1K+ daily calls at 99.5% uptime.
    • Designed fallback strategies, interruption handling and partial-response streaming for uninterrupted conversations.
    • Containerized and deployed AI services with Docker; used Redis for runtime performance and session consistency.
    • Delivered image segmentation pipelines for object and room-component detection in live production.
    • Advanced GenAI projects centred on deepfake technology and image-to-avatar transformation.
    • Worked with product, frontend and infra teams on scalable agent workflows and production-ready APIs.

  3. Machine Learning Engineer · Maknon Private Ltd.

    October 2023 – April 2024 (7 months) · Ahmedabad, Gujarat, India

    • Image segmentation to identify and categorize floors, walls and objects in images.
    • Fine-tuned segmentation models to improve accuracy and performance.
    • Room transformations with ControlNet for high-quality visual modifications.
    • Backend APIs to integrate the AI functionality.
    • Fine-tuned QLoRA-based transformer models to improve inference efficiency and reduce GPU memory use.
    • Optimized preprocessing and postprocessing to improve segmentation accuracy and runtime.

  4. AI Software Developer · SoluSoft Technologies

    October 2021 – October 2023 (2 years, 1 month) · Ahmedabad, Gujarat, India

    • Led development of Document AI, computer vision systems and LLM-powered chatbot solutions.
    • Document AI platform for enterprise clients, for classification and information extraction, using fine-tuned transformers with Django and FastAPI.
    • Built and fine-tuned transformer models for document classification, information extraction, sentiment analysis and dialogue summarization.
    • Barcode detection in documents with YOLO and OpenCV.
    • Car damage detection system with Detectron2 and CNNs.
    • Prototyped generative chatbots with LSTM sentiment analysis and BART dialogue summarization.
    • Led the team through to successful project completion.

Chapter III · वर्तमान

Pushing what generative AI can do.

Today, as a Senior Machine Learning Engineer at Bytes Technolab, I work on computer vision, Generative AI, AI automation and multi-agent workflows. Before that I built multi-agent LLM systems, real-time voice AI and image-to-avatar GenAI. These are the products I've shipped recently:

NoaAI

Multi-agent conversational and voice AI platform with centralized orchestration, strict phase isolation, RAG-based knowledge grounding, persistent conversation state, tool-based agent routing and low-latency streaming voice with interruption handling.

Dialora.ai

Real-time AI calling agent integrating STT, LLM reasoning and TTS over SIP, Twilio and PSTN, with fallback logic, cost-aware routing and conversation recovery.

dialora.ai

Ndeck.io

AI proposal and video generation: an LLM engine that turns structured inputs into complete technical proposals, with AI-generated diagrams, voice narration, visuals and lip-synced video.

ndeck.io

Looking ahead, I want to pioneer generative AI solutions for hard industry problems. If that's what you're working on, let's talk.

Chapter IV · गुरु

Even Arjuna needed a teacher.

On the chariot, Krishna taught and Arjuna listened. These are the places I learned.

  1. GLS Institute of Computer Technology (GLSICT-MBA)

    2015 – 2018

    Bachelor's Degree in Computer Applications.

  2. Udacity

    2020

    Nanodegree in Deep Learning.

Winner, Codiste Hackathon 2025

Epilogue · संवाद

Let's build the future of AI together.

Email me Read the blog