Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

AI/ML Engineer with 4+ years of production experience building and deploying machine learning systems, with a recent focus on LLMs and Generative AI. I’ve worked across LLM fine-tuning, RAG, agentic workflows, document intelligence, evaluation/observability, and scalable model serving using modern MLOps practices. I’m hands-on with PyTorch and data pipelines, and I’ve deployed solutions across GCP, AWS, and Databricks—designing reliable retraining, monitoring, and governance. I also build human-in-the-loop and confidence-based workflows to improve quality and reduce hallucinations in real-world deployments.…AI/ML Engineer with 4+ years of production experience building and deploying machine learning systems, with a recent focus on LLMs and Generative AI. I’ve worked across LLM fine-tuning, RAG, agentic workflows, document intelligence, evaluation/observability, and scalable model serving using modern MLOps practices. I’m hands-on with PyTorch and data pipelines, and I’ve deployed solutions across GCP, AWS, and Databricks—designing reliable retraining, monitoring, and governance. I also build human-in-the-loop and confidence-based workflows to improve quality and reduce hallucinations in real-world deployments.

Abhishek AI/ML Engineer+1





AI/ML Engineer with 4+ years of production experience building and deploying machine learning systems, with a recent focus on LLMs and Generative AI. I’ve worked across LLM fine-tuning, RAG, agentic workflows, document intelligence, evaluation/observability, and scalable model serving using modern MLOps practices. I’m hands-on with PyTorch and data pipelines, and I’ve deployed solutions across GCP, AWS, and Databricks—designing reliable retraining, monitoring, and governance. I also build human-in-the-loop and confidence-based workflows to improve quality and reduce hallucinations in real-world deployments.…AI/ML Engineer with 4+ years of production experience building and deploying machine learning systems, with a recent focus on LLMs and Generative AI. I’ve worked across LLM fine-tuning, RAG, agentic workflows, document intelligence, evaluation/observability, and scalable model serving using modern MLOps practices. I’m hands-on with PyTorch and data pipelines, and I’ve deployed solutions across GCP, AWS, and Databricks—designing reliable retraining, monitoring, and governance. I also build human-in-the-loop and confidence-based workflows to improve quality and reduce hallucinations in real-world deployments.

Available to hire

AI/ML Engineer with 4+ years of production experience building and deploying machine learning systems, with a recent focus on LLMs and Generative AI. I’ve worked across LLM fine-tuning, RAG, agentic workflows, document intelligence, evaluation/observability, and scalable model serving using modern MLOps practices.

I’m hands-on with PyTorch and data pipelines, and I’ve deployed solutions across GCP, AWS, and Databricks—designing reliable retraining, monitoring, and governance. I also build human-in-the-loop and confidence-based workflows to improve quality and reduce hallucinations in real-world deployments.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Language

English

Advanced

Telugu

Advanced

Work Experience

AI/ML Engineer (Contract) at PwC – Remote

April 1, 2025 - Present

Built an enterprise document intelligence platform for grounded Q&A, entity extraction, and document understanding across financial/audit/regulatory documents using a Databricks Medallion Lakehouse, fine-tuned LLMs, RAG, and agentic review workflows. Fine-tuned LLaMA-2 and Mistral with QLoRA/PEFT to match frontier model accuracy at lower GPU cost. Implemented production-grade RAG pipelines with LangChain and Databricks Vector Search (semantic chunking, hybrid search, cross-encoder reranking) over millions of documents. Integrated LayoutLMv3 for spatial-aware parsing of scanned documents. Designed multi-agent orchestration with planner/executor agents, function calling/tool use, state management, retries, and confidence-based escalation to human reviewers. Deployed LLM/classical models on Vertex AI and SageMaker with champion/challenger testing, shadow traffic, drift/quality monitoring, and automated rollback. Automated retraining via Cloud Composer and MLflow registry, and provisioned

AI/ML Engineer (Internship) at Blue Cross Blue Shield (BCBS) – Remote

September 1, 2024 - February 1, 2025

Developed a real-time risk scoring and decisioning platform combining gradient-boosted classifiers, graph neural networks, and ensemble anomaly detection for high-volume transactional events. Trained GBDT classifiers achieving 92% precision and 89% recall, reducing reviewer noise versus a legacy rules engine. Built PyTorch Geometric GNNs on entity-relationship graphs to uncover coordinated patterns missed by tabular models. Implemented ensemble anomaly detection to detect novel patterns between retrains. Prototyped an LLM-assisted triage tool using retrieval over historical decisions and policy documents to draft reviewer summaries. Built and managed a Feast feature store for offline/online serving to prevent train/serve skew. Added SHAP/LIME explainability and reason codes for audit readiness. Shipped scoring APIs with sub-100ms latency (FastAPI on GKE) and integrated into downstream workflows. Led migration from Redshift to Snowflake and rebuilt batch ETL with Airflow/Glue/Lambda and

Data Scientist at Jotter.ai – Hyderabad, India

September 1, 2021 - July 1, 2023

Owned an end-to-end automated visual catalogue generation platform for large-scale SKU automation. Trained ResNet50 and EfficientNet classifiers with custom multi-task heads for multi-label attribute tagging (category, color, material, pattern), replacing manual labeling. Built Blender-based 2D-to-3D reconstruction to generate textured 3D product models and photorealistic multi-angle catalogue views, including synthetic data for augmentation. Implemented Mask R-CNN and U-Net segmentation with post-processing for reflective/transparent surfaces. Created visual similarity search using Siamese networks with triplet loss to produce embeddings for nearest-neighbor retrieval and deduplication. Deployed model serving and image processing via FastAPI on GKE with GPU autoscaling; optimized inference using ONNX and TensorRT. Built async pipelines with Celery/Redis for scalable parallel image processing, reducing time-to-catalogue by 60%.

Data Scientist (Internship) at Flutura Decision Sciences & Analytics – Remote, India

August 1, 2020 - August 1, 2021

Built predictive maintenance models for industrial mining trucks using time series analysis over multivariate sensor telemetry. Engineered temporal features from raw sensor streams to capture degradation patterns and failure signatures. Trained classification models to predict equipment failures before they occur, reducing unplanned downtime risk.