Hi, I'm Dibyajyoti Mohanty

AI/ML Engineer | Agentic AI | Conversational AI | Multimodal AI

Building production-grade AI systems beyond simple wrappers — spanning multi-agent orchestration, RAG, real-time speech, LLM infrastructure, multimodal AI, and GPU-optimized inference.

Get In Touch Download Resume

About Me

I’m an AI/ML Engineer focused on designing and building production-grade AI systems, with experience across Generative AI, Agentic AI, Conversational AI, Speech, NLP, Computer Vision, and multimodal systems. My work goes beyond integrating model APIs — I engineer the systems around them, including retrieval pipelines, agent orchestration, inference optimization, real-time communication, reliability, and scalable deployment.

I’ve built real-time speech systems using ASR, TTS, voice cloning, speaker diarization, and VAD; developed RAG and multi-agent workflows with dynamic tool routing and memory; and worked with Vision-Language Models, semantic segmentation, and multimodal AI. I also focus heavily on performance engineering, including GPU utilization, multi-GPU scheduling, inference latency, memory optimization, and high-throughput AI pipelines.

My engineering approach is centered around taking AI systems from research to production — making them reliable, scalable, observable, and maintainable. I enjoy solving the difficult engineering problems behind AI products, whether that means reducing latency, improving inference efficiency, designing robust LLM infrastructure, or building intelligent systems that can operate reliably at scale.

85% QA Workload Reduction
75% HR Ticket Deflection
80% Faster Provisioning

Professional Experience

SDE – AI Engineer

Movius (Contract via Canarys: Feb 2025 – Jan 2026 | Full-time: Feb 2026 – Present)

Feb 2025 - Present

  • Built ASR-OTT, a real-time communication application for people with speech and hearing impairments using Faster-Whisper (ASR), StyleTTS2 (TTS), and Mistral-70B (summarization)
  • Implemented voice cloning in TTS pipeline and extensive audio pre/post-processing for accurate transcriptions
  • Designed agentic workflow with dynamic prompting for conversation summaries
  • Built context-aware reply system using RAG with NVIDIA Nemotron and vector databases
  • Optimized GPU utilization using round-robin scheduling across AI services
  • Managed GPU AI infrastructure: CUDA, cuDNN, NVIDIA drivers, PyTorch, TensorFlow compatibility

Generative AI Engineer

ATG

Jan 2024 - Sep 2024

  • Led development of two end-to-end AI solutions as Lead AI Engineer using Python, ML, LLMs, NLP, Docker, and MLOps
  • Designed and implemented RAG pipelines to improve contextual understanding and response accuracy
  • Developed AI-powered Sales Call Analysis platform with speaker diarization (Deepgram), transcription, and LLM-based performance evaluation using Llama-3

AI/ML Consultant

Freelance

Feb 2023 - Dec 2023

  • Delivered end-to-end AI and Generative AI solutions using Python, LLMs, NLP, Computer Vision, RAG, and MLOps
  • Developed AI conversational systems with real-time audio analysis integrating Gemini, Telegram Bot API, ASR, TTS
  • Built AI applications with OpenAI DALL·E, Flask, MongoDB, and U-Net medical image segmentation models
  • Fine-tuned Qwen 3.5 Vision-Language Models using QLoRA with Unsloth framework
  • Implemented Human-in-the-Loop reinforcement learning workflows to improve VLM performance

Programmer Analyst

Cognizant

Aug 2022 - Jan 2023

  • Performed data analysis, transformation, and pre-processing using Python and MySQL for ML and analytics
  • Developed and optimized data pipelines, improving data quality and processing efficiency
  • Collaborated with cross-functional teams to deliver actionable insights for data-driven decision-making

Software Developer

Stack System

Feb 2020 - Aug 2021

  • Developed backend features for Flask-based web application with scalable RESTful APIs
  • Optimized application performance using Python and best practices
  • Mentored graduate students in Python programming and software development

Featured Articles & Writing

Technical guides, RAG system architectures, and AI engineering insights published on Medium

Medium 6 min

Creating Scalable RAG Systems with LangChain, Neo4j, and Advanced LLMs

Published in GoPenAI

Combining Knowledge Graphs (Neo4j) with LangChain and Large Language Models to construct Graph-based RAG systems for structured context retrieval.

Graph RAG Neo4j LangChain LLMs
Read on Medium
Medium 7 min

AI at Work: Automating Meeting Transcriptions & Performance Reviews

Automating workplace productivity by integrating AssemblyAI speech-to-text transcriptions with Multimodal Retrieval-Augmented Generation for performance reviews.

Multimodal RAG AssemblyAI Speech AI LLM Pipeline
Read on Medium
Medium 5 min

Building a RAG Chatbot with Qdrant, LangChain, and Streamlit

Conversational RAG application using Qdrant vector database for high-performance similarity search, LangChain for memory & retrieval, and Streamlit UI.

Qdrant Vector DB LangChain Streamlit RAG
Read on Medium
Medium 8 min

A Comprehensive Guide to Data Scaling Techniques

In-depth technical breakdown of data scaling methodologies (Normalization, Standardization, Robust Scaling) in ML feature engineering.

Data Scaling ML Engineering Preprocessing Python
Read on Medium

Skills & Technologies

Languages & Core Libraries

Python (5+ years) Pandas NumPy Matplotlib Seaborn Scikit-learn C++

AI/ML Frameworks

PyTorch TensorFlow Keras Scikit-learn Hugging Face Transformers LangChain LangGraph NVIDIA NeMo BERT Sentence Transformers OpenAI Google AI

Machine Learning

Supervised Learning Unsupervised Learning Reinforcement Learning Anomaly Detection Transfer Learning Ensemble Methods Dimensionality Reduction Model Selection Cross-Validation Hyperparameter Tuning Feature Engineering Data Preprocessing

Deep Learning

CNN R-CNN YOLO Deep SORT RNN LSTM Bi-LSTM Transformers LLMs Autoencoders GANs VAEs Seq2Seq Attention Mechanisms Multi head Attention Self-Attention Positional Encoding Speculative Decoding

Generative AI & LLMs

RAG Agentic RAG Multimodal RAG Knowledge Graph RAG Conversational Memory Prompt Engineering AI Agents Fine-tuning LLM Evaluation LLM Deployment

Computer Vision

Image Classification Object Detection Semantic Segmentation Facial Recognition OpenCV torchvision PaddleOCR U-Net YOLO VLM Multi Modality

Speech & Audio AI

ASR TTS Voice Cloning Speaker Diarization Speech-to-Text Text-to-Speech Faster-Whisper Deepgram VAD Audio Classification Audio Streaming

Databases

SQL MySQL MongoDB Neo4j GraphDB Qdrant ChromaDB FAISS PostgreSQL Redis Kafka Elasticsearch MariaDB BM25 VectorDB

GPU & AI Infrastructure

CUDA cuDNN NVIDIA Drivers GPU Optimization AI Inference Model Deployment Dependency Management Multi-GPU Scheduling NVIDIA Nsight

Version Control & DevOps

Git GitHub Bitbucket Docker GitHub Actions CI/CD Linux Kubernetes Helm Grafana ELK Stack

Cloud & MLOps

AWS SageMaker AWS ECS AWS API Gateway GCP MLflow Airflow

Web Frameworks & APIs

Flask FastAPI Streamlit REST APIs

Development Tools

VS Code Cursor Jupyter Notebook Postman Conda Git

AI Domains

Computer Vision NLP LLMs Generative AI RAG Systems Agentic AI Conversational AI Multi-Modality

Featured Projects

AI-Powered NPC System with Dynamic Personality

Multi-layered AI agent system featuring persona core, perception layer, decision engine, and LLM integration (Groq/Llama 3.3 70B) for real-time dialogue. Implements dual-memory system with short-term and long-term storage, relationship trust scores, emergent social dynamics, NPC-to-NPC conversations, and dynamic world events.

LangChain LangGraph Llama 3.3 AI Agents Memory Architecture

Multi-Agent AI Assistant with Long-Term Memory

Multi-agent conversational AI system using LangChain, LangGraph, Groq Llama-3, and Flask/React, supporting threaded conversations and dynamic context switching. Features agent orchestration, tool routing, and memory management to enable collaborative reasoning and context-aware dialogue.

LangChain LangGraph Flask React Groq

Enterprise Conversation Intelligence Platform

AI-powered conversation intelligence platform analyzing client–business analyst interactions using ASR, speaker diarization, and RAG. Built retrieval pipeline with Google Embedding-Text-004, ChromaDB, and semantic chunking. Leveraged Deepgram Nova-2 for diarization and Groq Llama-3 for contextual evaluation and performance scoring.

RAG ChromaDB Deepgram Llama-3 NLP

AI Sales Call Analysis Platform

AI-powered platform processing customer-sales conversation audio with speaker diarization to identify salesperson's speech, transcription, and LLM-based evaluation of sales performance, response quality, and communication effectiveness using Deepgram and Llama-3.

Deepgram Llama-3 Speaker Diarization NLP

AI Tattoo Generator with Advanced Vision

Full-stack AI application for personalized tattoo design generation using DALL·E 3, Stable Diffusion XL, and ControlNet. Features dual-mode generation: text-to-image with style transfer and image-based generation with MediaPipe pose estimation and DeepLabv3+ semantic segmentation for precise body part detection. Implements LoRA fine-tuning for style consistency, prompt engineering with negative prompts, and real-time preview with latent diffusion optimization.

DALL·E 3 Stable Diffusion XL ControlNet MediaPipe DeepLabv3+ LoRA React Flask MongoDB

Education

B.Sc. Information Science and Telecommunication

Ravenshaw University

2022

Focus: Computer Science, Data Science, and Telecommunication Systems

Strong foundation in algorithms, data structures, machine learning, and communication technologies. Developed skills in programming, statistical analysis, and problem-solving that directly translate to AI/ML engineering.

Get In Touch

I'm always open to discussing new projects, creative ideas, or opportunities to be part of your vision.

Email

dibyamohanty069@gmail.com

LinkedIn

Connect with me

GitHub

Check out my work