Less
More
BUILDING AI SYSTEMS THAT SCALE
Integrated Neo4j for a live knowledge graph and Milvus for vector search. Deployed microservices via FastAPI and NGINX on GCP & AWS.
Implemented RAG pipelines with context persistence and chain-of-thought reasoning — enabling real-time, accurate responses and reducing hallucinations.
Performance Real-time vector retrieval, dynamic prompt routing, and multi-agent orchestration ensure high-fidelity outputs and low latency across all system layers.
Engineering Agentic LLM Systems
Built Agentic AI Chatbot, a multi-agent mental health chatbot with 20+ agents orchestrated across 100+ nodes using Kafka and Redis Streams.
Real-Time Speech AI & TTS
Fine-tuned Whisper, Indic Conformer, and TTS models on A100 GPUs. Developed real-time transcription services with low-latency, high-accuracy results.
Retrieval-Augmented Generation and Vector Search
Designed and implemented production-grade RAG pipelines powering real-time, context-aware conversational AI. Leveraged semantic search, embeddings, and multi-agent orchestration for dynamic and accurate information retrieval.
Vector DB
Milvus, Qdrant, FAISS
Frameworks
LangChain, LlamaIndex, LangGraph, mem0
LLMs
Meta Llama, Mistral, DeciLM, OpenAI, Claude, Gemini, Qwen
Infra
Kafka, Redis, FastAPI, GCP, Docker, AWS
83% Faster Retrieval
Optimized field generation time (30 → 5 mins) through async pipelines & message brokers.
Dynamic Vector APIs
Real-time indexing, semantic search & adaptive embeddings at scale.
20+ AI Agents
Orchestrated across 100+ microservices with LangGraph, Redis & Kafka.
Professional Journey
My journey through education, research, and professional experience in AI and machine learning.
Education
Education
B.Tech in Computer Science and Engineering
Techno Engineering College, Banipur
2019 – 2023
Focused on AI, ML, and Software Engineering; completed projects in stock prediction, plant disease detection, and emotion analysis.
Central Board of Secondary Education (AISSCE)
Kendriya Vidyalaya, Mumbai
2018 – 2019
Central Board of Secondary Education (SSC)
Kendriya Vidyalaya, Mumbai
2016 – 2017
Research
Research
Published research in AI-driven mental health solutions, machine learning applications, and healthcare technologies.
Chatbot-Enhanced Mental Health First Aid in Corporate Settings
S Banerjee, A Agarwal, AK Bar • American Journal of Software Engineering and Applications, Vol. 12(1), 2024
Boosting Workplace Well-being with Mental Health Chatbot
S Banerjee, A Agarwal, P Ghosh, AK Bar • American Journal of Artificial Intelligence, Vol. 8(1), 2024
Securing Well-being: Security Protocols in AI-driven Mental Health Chatbots
S Banerjee, A Agarwal, AK Bar • American Journal of Computer Science and Technology, Vol. 7(1), 2024
Disease Detection for Herbal Plants Using ResNet Algorithm
A Rout, AK Bar, R Sarkar, AK Chaudhuri • AI and IoT Technology and Applications for Smart Healthcare Systems, 2024
Emotica.AI – A Customer Feedback System using AI
AK Bar, AK Chaudhuri • International Research Journal on Advanced Science Hub, Vol. 5(03), 2023
Cryptojacking Detection Using Genetic Search Algorithm
AK Bar, A Rout, AK Bar • International Research Journal on Advanced Science Hub, Vol. 5(04), 2023
Stock Market Prediction using Machine Learning Algorithm
A Rout, AK Bar, SP Saha • International Journal of Advanced Research in Computer and Communication, 2022
Experience
Experience
Applied AI Scientist & Founding Member
United We Care X Shunya Labs AI
06/2025 – Present
Fine-tuned large-scale Whisper, Indic Conformer, and TTS models using LoRA (PEFT) and Seq2Seq Trainer frameworks on A100 GPUs
Developed and deployed real-time speech transcription service handling 250+ concurrent requests
Fine-tuned Whisper ASR model (shunyalabs/pingala-v1-en-verbatim) achieving 2.94% WER, surpassing industry benchmarks
Contributed to Speech AI and Voice AI research, including dataset preparation, model optimization, and GPU utilization
AI Engineer
United We Care
04/2024 – 06/2025
Built "Stella 2.0", a multi-agent mental health chatbot ecosystem with 15 autonomous agents across 60+ nodes
Developed Clinical Copilot for healthcare professionals using advanced prompt engineering and Chain-of-Thought methods
Managed deployment of microservices on AWS & GCP, using FastAPI and NGINX for scalable and reliable APIs
Implemented RAG & vector search pipelines with Redis and Milvus, reducing field generation time by 83%
Deployed multiple HuggingFace LLMs (Meta-Llama-3/3.1, DeciLM-7B, Mistral-7B) for production NLP tasks
AI Research Intern
United We Care
12/2023 – 03/2024
Conducted research on AI-driven mental health solutions, resulting in three international publications
Built POCs with Metarank, Docker, Kafka, Redis, and vector databases (Milvus, Qdrant, FAISS, ChromaDB)
Experimented with LLMs (Mixtral-8x7B, Meta-Llama-3/3.1, DeciLM-7B) using LangChain, LlamaIndex, and LangFlow
About Me
I'm an AI Engineer specializing in Speech AI, Voice AI, and Large Language Models (LLMs) — passionate about building scalable, production-grade AI systems that bridge research and real-world applications. My work focuses on fine-tuning large-scale ASR, TTS, and LLM models using LoRA, PEFT, and distributed training frameworks to achieve SoTA (state-of-the-art) performance.
I have led the development of real-time speech transcription and conversational AI systems, optimizing for low-latency inference, high concurrency, and GPU efficiency. I've built agentic AI ecosystems, retrieval-augmented generation (RAG) pipelines, and cloud-deployed AI microservices on AWS and GCP, integrating tools like LangChain, LangGraph, Redis, Neo4j, and Milvus to deliver reliable, intelligent automation at scale.
Driven by a balance of research rigor and engineering execution, I enjoy pushing the boundaries of what's possible with AI — from speech recognition to intelligent assistants — while actively contributing to open-source and research communities.