Open to opportunities

Aakash Srivastava

@aakashsrivastava

Message

Research Scientist building edge speech and LLM systems.

India

Message

What I'm looking for

I’m looking to build production-grade, low-latency AI for speech and LLMs—especially edge/on-device deployments—where I can own the pipeline end-to-end, optimize performance, and ship reliable systems that work in real environments.

I’m a Research Scientist focused on training, fine-tuning, and deploying compact large language models and multilingual speech systems for low-latency, on-device performance. I work end-to-end—from data curation and optimization to production inference stacks—using GPU-accelerated PyTorch workflows and memory-aware training.

At Bosch, I delivered production-ready speech and voice authentication capabilities, including an edge-based Speech-to-Text pipeline and a custom wake word detection system for multiple languages. I also build production C++/JNI inference for ARM CPUs and Android integration, and I’ve developed multilingual Encoder-Decoder ASR and speech-to-text translation pipelines to power real-time, LLM-enabled applications.

Experience

Work history, roles, and key accomplishments

Current

Research Scientist

Current

Cerence AI

Jun 2025 - Present (1 year 1 month)

Trains and fine-tunes compact LLMs for GPU-accelerated workflows and deploys low-latency speech/LLM inference for ARM-based edge devices. Builds multilingual ASR and end-to-end speech pipelines integrating ASR with Audio Speech Translation for speech-to-text translation.

PyTorch Mixed Precision Training Fine Tuning C++KV Cache Management Streaming Inference Android Encoder Decoder ASR Multilingual ASR Audio Speech Translation Knowledge Distillation RAG

Software Engineer

Bosch Global Software Technologies Pvt Ltd

Jul 2023 - Jun 2025 (1 year 11 months)

Develops production edge AI systems for speech, wake-word detection, and voice authentication, including cloud/edge deployment and Android integration. Builds and tunes speech-to-text models for noisy automotive environments and manages large-scale GPU training and CI/CD pipelines.

Project Trainee

Bosch Global Software Technologies Pvt Ltd

Nov 2022 - Apr 2023 (5 months)

Builds an offline voice assistant pipeline combining Automatic Spoken Language Identification, speech recognition, and task automation for Indian market use cases. Develops an ML pipeline to convert Figma flow images into intermediate JSON for requirement engineering and creates UI/UX test cases for app verification.

Language Identification Speech Recognition Speech to Text Task Automation Machine Learning Pipelines Figma UI Testing App Testing data preprocessing Neural Networks

Education

Degrees, certifications, and relevant coursework

Bharati Vidyapeeth (Deemed to be University) College of Engineering

Bachelor of Technology, Computer Engineering

2019 - 2023

Grade: CGPA: 9.12

Earned a Bachelor of Technology in Computer Engineering from Bharati Vidyapeeth (Deemed to be University), College of Engineering, Pune, completing the program in 2023.