Open to opportunities

abhinav sharma

@abhinavsharma6

Message

AI & ML engineer focused on building scalable inference systems and production-ready generative models.

India

Message

What I'm looking for

I am seeking roles where I can build production ML systems or train large models, working on scalable inference, robust validation, and impactful AI products.

I am an AI and machine learning engineer with hands-on experience building end-to-end ML pipelines, production inference services, and large-scale model training. I have delivered systems that generate thousands of real-time personalized outputs and designed robust model training workflows for stability and performance.

My work spans generative transformers, real-time vision analytics, and ensemble-based fraud detection, with accomplishments including a 26M-parameter transformer pretrained on hundreds of millions of tokens, a low-latency FastAPI inference stack producing 10,000+ personalized reports, and a leakage-resistant XGBoost stacking framework achieving strong ROC-AUC and PR-AUC metrics.

I prioritize production readiness, low-latency scalable services, and rigorous validation practices, and I enjoy reducing manual effort through automation and improving model reliability through principled engineering.

Experience

Work history, roles, and key accomplishments

ML Researcher

Personal Projects

Jan 2026 - Feb 2026 (1 month)

Developed and trained a 26M-parameter GPT-2 style Transformer achieving 49.5 validation perplexity, engineered a 20K-token BPE tokenizer and a mixed-precision GPU training pipeline to enable stable large-scale training.

PyTorch CUDA AMP Transformers Tokenization Gradient Clipping Learning Rate Scheduling

Computer Vision Engineer

VisionGuard

Dec 2025 - Jan 2026 (1 month)

Built a real-time video analytics pipeline processing streams at 25 FPS for safety violation detection, implemented temporal consistency modeling to reduce false positives by 30%, and designed asynchronous FastAPI microservices for low-latency inference and audit logging.

OpenCV fastAPI Video Processing Computer Vision

Machine Learning Engineer

Credit Card Fraud Detection

Sep 2025 - Oct 2025 (1 month)

Built a leakage-resistant stacked XGBoost ensemble with temporal validation and optimized thresholds to improve minority-class recall by 18%, achieving 0.99 ROC-AUC, 0.80 PR-AUC, and 0.81 F1 under extreme imbalance.

XGBoost scikit learn Imbalanced Learning Model Evaluation

AI Intern

PhotoGPT AI

May 2024 - Jul 2024 (2 months)

Architected an end-to-end ML skin analysis pipeline producing structured dermatological reports and rule/database-backed recommendations; deployed FastAPI inference services that generated 10,000+ personalized reports with sub-second latency and automated product-matching that reduced manual intervention by 80%.