Pyla Sandeep

About Pyla Sandeep

I’m Pyla Sandeep, a B.Tech graduate in Information Technology from JNTU Gurajada, Vizianagaram, with a strong interest in Artificial Intelligence, Machine Learning, Deep Learning, and Full Stack Development.Currently, I’m working as an AI/ML Engineer Intern at Chisel AI Solutions, where I focus on data preparation, preprocessing, model fine-tuning, and training for speech technologies like Text-to-Speech (TTS) and Speech-to-Text (STT).I’ve gained hands-on experience through internships and projects, including:1) Developed a Breast Cancer Detection model with 98% accuracy at NIT Warangal2) Worked with Large Language Models (LLMs) for question-answering systems at Veritometrics3) Building impactful solutions like an Indian Sign Language Translator and an AI-powered Speech Enhancement toolI’m passionate about using AI to solve real-world problems and create solutions that bring social impact. Beyond AI, I also enjoy full stack web development and exploring emerging technologies.Always open to collaboration, learning opportunities, and meaningful discussions in the field of AI and technology. Let’s connect!

Experience

1. Software Engineer (Ai Ml)

   Chisel Solutions

   Nov 2025 — Present

   Worked on end-to-end AI systems across Speech-to-Text (STT), Document AI, and ML infrastructure, with a strong focus on production readiness and scalability.Built multilingual ASR pipelines (10–12 Indian languages) involving data preparation, NeMo compatible manifest creation, tokenizer training, and from-scratch training of FastConformer + RNNT and Whisper based models.Designed and deployed real time and batch STT systems, including Faster-Whisper optimization, WebSocket-based audio streaming with VAD, and GPU inference using NVIDIA Triton and NVIDIA Riva.Worked extensively on ML systems engineering, covering Linux server setup, CUDA/cuDNN configuration, environment management with pyenv, Docker based deployments, and efficient dataset transfer and compression.Developed a synthetic data generation platform for U.S. IRS tax forms (W-2, 1040, 1099 family), aligned with IRS 2026 updates, using schema driven design and compliance aware validation.Implemented high performance streaming CSV pipelines with O(1) memory usage, reproducibility, and per-field distortion tracking for auditability.Built pixel accurate PDF and image rendering pipelines, converting CSV data into IRS style forms and OCR ready datasets.Applied multi tier data and image distortions (formatting noise, rounding, truncation, blur, skew, lighting, compression) to simulate real world document variability.

Education

• Tirumala Junior College

  Intermediate , MPC

• JNTU Gurajada Vizianagaram

  Bachelor of Technology - BTech, Information Technology