Looking for a job

VAIBHAV SINGH

@vaibhavsingh16

Message

I build production-ready computer vision, speech, and LLM systems.

India

Message

What I'm looking for

I'm looking to build production ML and AI systems where I can work on computer vision, speech, LLMs, and model optimization—from experimentation through deployment and measurable product impact.

At AfterShoot, I built ColorNet to improve color-model accuracy by ~20% and led a RAW image denoising model from dataset creation through evaluation, reaching 44 dB PSNR while being 3x smaller and ~15x faster than Restormer. I also built local training, inference, and benchmarking pipelines and personalization models using XGBoost.

Previously, I built AI chatbots, voice assistants, agentic sales workflows, and real-time transcription systems using Python, PyTorch, FastAPI, BERT, Whisper, NVIDIANeMo, and LLMs. I enjoy taking ML experiments into practical systems across computer vision, speech, and generative AI.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer - I

Current

AfterShoot

Oct 2025 - Present (11 months)

Designed and built ColorNet, a custom architecture improving model accuracy by ~20% and consistency by ~23%. Took ownership of a RAW image denoising model end-to-end, achieving ~44 dB PSNR while being 3x smaller and ~15x faster than Restormer.

Python PyTorch TensorFlow MLOps Statistics Computer Vision Image Processing XGBoost Model Quantization Onnx Docker MLFlow

AI Intern

BASALAI

Jun 2025 - Oct 2025 (4 months)

Built a POC for an autonomous AI SDR, integrating Apollo.io, Lusha, and Coresignal into a unified lead generation pipeline. Worked with Newsweek on automating news and article refinement pipelines, and built custom chatbots using Copilot Studio and PowerAutomate.

Agentic AI OpenAI LangGraph Pandas CoPilot Studio Power Automate Whisper Docker

Data Science Intern

OriServe

Jul 2024 - May 2025 (10 months)

Worked on an AI chatbot and voice assistant using BERT, Pyannote, and NVIDIA NeMo, improving user engagement and operational efficiency by ~20%. Designed and deployed scalable FastAPI services and engineered pipelines for TTS and STT datasets.