About Me

Professional Journey

I'm Tarun, a Data Scientist at Jio Platforms Ltd, where I architect and deploy enterprise-scale AI solutions that transform business operations. My Electrical Engineering degree from IIT Bombay gave me the mathematical grounding, which helps me specialize in optimizing AI models for complex business problems.

My work focuses on building robust, scalable AI systems across computer vision, speech processing, natural language processing and GenAI agentic workflows. I leverage cutting-edge technologies like PyTorch, LangChain, Restful APIs, Containerization, and Kubernetes to create high-performance solutions that deliver real-time responses and scale to an enterprise level.

Career Timeline

2023 - Present
Data Scientist
Jio Platforms Ltd

Building and deploying enterprise AI solutions for speech, vision NLP applications and GenAI agentic workflows, and building and deploying scalable and efficient AI models for production environments.

2019 - 2023
Bachelor of Technology
IIT Bombay - Electrical Engineering

Specialized in signal processing with focus on image and speech processing applications, building a strong foundation in digital signal processing that transitioned into ML/AI.


Featured AI Project

Finance-Llama-8B

A fine-tuned Llama 3.1 8B model specialized in financial reasoning, question answering, and multi-turn conversations. Trained on over 500,000 financial entries spanning QA, sentiment analysis, and conversational AI.

Key Features: Financial specialization, multi-turn conversations, 73% CFA Level 1 performance, 500k+ training entries.

📈 3k+ downloads per month on Hugging Face and 1k+ downloads per month on Ollama

Core Competencies

AI/ML Technologies

  • PyTorch & Distributed Training
  • CUDA & GPU Optimization
  • Kubernetes & Docker
  • LangChain & LangGraph
  • LLM Fine-tuning & RAG Systems
  • MCP Servers & Agentic Workflows
  • Computer Vision (OCR, Object Detection)
  • Speech Processing (TTS, ASR)

Infrastructure & Deployment

  • NVIDIA MIG & Multi-GPU Clusters
  • Flask & FastAPI Development
  • Redis & Celery for Async Processing
  • Nginx & Gunicorn Deployment
  • MLOps & Model Serving
  • Monitoring & Logging (ELK Stack)
  • Containerization & Kubernetes
  • Restful APIs

Key Achievements

  • Enterprise AI Chatbot: Built HR Assistant serving 10k+ employees with Graph RAG pipeline, improving retrieval accuracy by 30% and response personalization by 20%
  • Real-Time TTS Engine: Delivered cross-lingual TTS APIs achieving 0.2 RTF with 100+ concurrent streams on NVIDIA MIG instances at 90% GPU utilization
  • Multilingual OCR System: Fine-tuned PaddleClas PULC for 10 languages achieving 93.54% F1 score, with 25% accuracy improvement through adaptive model routing
  • Computer Vision Pipeline: Fine-tuned RT-DETR (AP=0.884) and DINOv2 (93.51% F1) for object & landmark detection with scalable Kubernetes deployment handling 4x peak loads
  • Distributed Training: Orchestrated PyTorch DDP on 4 NVIDIA A100 GPUs reducing training time by 40% and deployed with Gunicorn/Nginx cutting API errors to 0.5%
Tarun

Connect

I'm always interested in connecting with fellow AI enthusiasts, researchers, and professionals. Feel free to reach out if you'd like to discuss AI/ML projects, collaborate on interesting problems, or simply connect over shared interests in technology and sports.