Aditya Shravan K.

Key Skills

Software

AWS SageMaker

Other

Top Subject Matter

Healthcare datasets (provider encounter, revenue, immunisation)

Clinical AI for healthcare summarization and multimodal document understanding

Clinical document information extraction (medications, diagnoses, lab results)

Top Data Types

Text

Document

Image

Top Task Types

Text Summarization

Text Generation

Prompt + Response Writing (SFT)

Classification

Entity (NER) Classification

Data Collection

Freelancer Overview

Software AI Engineer Intern, LucidAct Health Inc (AI/GenAI clinical evaluation and extraction workflows). Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include AWS SageMaker and Other. Education includes Master of Science, Rowan University (2025). AI-training focus includes data types such as Medical, DICOM, and Computer Code and labeling workflows including Text Summarization, Data Collection, and Classification.

Labeling Experience

Software AI Engineer Intern, LucidAct Health Inc (AI/GenAI clinical evaluation and extraction workflows)

AWS SageMakerText Summarization

Developed and benchmarked LLMs for clinical summarization and multimodal vision tasks using an ELO-style ranking evaluation harness. Built a production GenAI clinical platform that uses physician feedback to improve model outputs for clinical decision support. Implemented an image-to-text extraction pipeline to derive structured clinical fields from medical documents. • Evaluation of clinical summarization quality using ranking-based methodology. • Multimodal image-to-text extraction from clinical documents. • MRN-scoped session management to reduce redundant context and input tokens. • Prompt microservice with schema-driven templates for deploy-less prompt updates.

2025 - 2025

Machine Learning Researcher - Collaborative Systems Inc

ImageClassification

Served as a Machine Learning Researcher focused on blockchain risk detection using semi-supervised anomaly detection techniques. Developed ML pipelines in Python with Web3 data sources and trained multiple models to classify and identify high-risk wallet behavior. Validated model findings against known exploit references to support safer protocol selection and downstream integration decisions. • Built a semi-supervised blockchain ML pipeline using Python, Web3.py, PyOD, and LightGBM for 10K+ on-chain records • Trained and compared five ML models for wallet classification, selecting the best-performing approach on imbalanced fraud data • Engineered wallet-level anomaly detection features and evaluated them using metrics such as F1-score • Confirmed flagged wallets against Ethereum Foundation exploit databases to identify fraud signatures for DEX integration decisions

2025 - 2025

Software AI Engineer Intern - Lucidact Health Inc

ImageClassificationEntity (NER) Classification

Worked as a Software AI Engineer Intern delivering production-ready generative AI for clinical summarization and multimodal document understanding. Built and evaluated LLM systems on AWS using Bedrock models, ranking-based evaluation, and serverless architectures. Collaborated with domain stakeholders to improve model outputs using physician feedback and streamlined prompt management for repeatable deployments. • Designed an LLM evaluation harness benchmarking models on clinical summarization and multimodal vision • Shipped a production GenAI clinical platform on AWS Lambda and Bedrock using Claude Sonnet 4 • Deployed a multimodal image-to-text pipeline extracting medications, diagnoses, and lab results • Architected MRN-scoped Bedrock session management and DB-backed prompt microservices to reduce token usage and enable deploy-less prompt updates

2025 - 2025

Machine Learning Researcher — Blockchain Risk Detection (semi-supervised ML pipeline)

Classification

Created a semi-supervised machine learning pipeline to detect blockchain wallet risk using anomaly detection and classification models. Engineered wallet-level anomaly features from on-chain data and trained multiple models for fraud-related classification. Validated model-flagged entities against known exploit references to confirm signatures relevant for safe protocol selection. • Processed 10K+ on-chain records and engineered 20+ wallet anomaly detection features. • Trained Isolation Forest, LOF, AutoEncoder, and LightGBM models for wallet classification. • Achieved best F1 with LightGBM on imbalanced fraud data. • Verified flagged wallets against Ethereum Foundation exploit databases to confirm DeFi fraud signatures.

2025 - 2025

Software AI Engineer Intern — LucidAct Health Inc (Prompt microservices & template-driven generation)

TextPrompt + Response Writing (SFT)

Built schema-driven prompt templates and a DB-backed prompt microservice to enable controlled generation behavior across clinical workflows. Decoupled GenAI prompting logic from application code to allow deploy-less prompt updates while maintaining consistent structured outputs. Supported reliable function calling and workflow execution through templated prompt design. • Designed DB-backed prompt microservice with schema-driven templates. • Decoupled GenAI logic from application code for safer iteration. • Enabled deploy-less prompt updates across clinical workflow stages. • Ensured prompts could be updated without redeploying services.

2025 - 2025

Show more

Education

Work History