Abhishek Mathada

About Abhishek Mathada

I\'m a Machine Learning Engineer with a passion for coding (Python) and a knack for…

Experience

1. Machine Learning Engineer

   Jio

   Jan 2022 — Present · Navi Mumbai, IN

   Omni All-in-One Assistant (In Progress) – Built an advanced LLM powered multi-agent orchestrator integrating MCP tools & agents. Supports sequential/parallel execution, multimodal inputs (text+images), image generation, and personalized memory service.Skills: Spanner DB, Qdrant, Google KMS, mem0, openai-agents-python, AWS Agent Squad, LiteLLM, Langfuse. HelloJio LLM Chatbot (Production | 4–5 Lakh Daily Users) – Designed a multilingual RAG-based chatbot in the MyJio app using models (Llama 3.3, GPT-4, GPT-4o, GPT-4o Mini, GPT-3.5, Gemini Flash 2.0). Features include search, session management, recent history, feedback, audio mode, regeneration, and low-latency caching.Skills: Prompt engineering, retrieval, caching, WebSocket, Monstache, FastAPI, Azure Cloud, Azure OpenAI, LiteLLM. EVA (Engage Virtual Assistant) – Created a chatbot platform enabling clients to build bots using intent classification (LR, LSTM) & entity recognition (spaCy, regex). Supports custom datasets with mapped responses.Skills: spaCy, ML models, regex, Celery, MongoDB, Elasticsearch. HelloJio Voice Assistant (Legacy) – Developed a voice assistant (Android/iOS) on EVA platform using ML for intent classification & entity recognition, integrated with ATP workflows for Jio troubleshooting. LLM GPU Benchmarking (Ongoing) – Evaluating GPUs (Nvidia H200, AMD MI300X) across runtimes (vLLM, TensorRT, SGLang). Focus on TFLOPS, bandwidth, ISL/OSL, concurrency, and metrics like TTFT & TPOT.Skills: LLM runtimes, PyTorch profiling, GPU optimization, GEMM benchmarking.