Open to opportunities

Abhinav gupta

@abhinavgupta6

Message

Data Science Intern building production AIML solutions for health fintech.

India

Message

What I'm looking for

I want to work on production ML/LLM systems with measurable impact—medical/document intelligence, real-time detection, and strong observability. I’m excited by teams that value experimentation, optimization, and reliable deployment pipelines.

I’m a Data Science Intern (Computer Science) focused on turning ML into reliable, production-grade systems. In my current role, I built ML/DL solutions that deliver strong performance on cataract detection and OPD prescription extraction.

I design end-to-end pipelines—from large-scale data processing to model deployment—using PyTorch/TensorFlow/Scikit-learn and modern optimization (Optuna, PEFT/LoRA/QLoRA). I’ve also engineered a claims IPD pipeline that processes 1.2 lakh claims/month, cutting costs by INR 14L/month and manual effort by 61%.

I enjoy pushing LLM and document AI forward with evaluation frameworks, LangChain/LangGraph workflows, and observability (LangFuse, MLflow). Through projects like real-time claims fraud detection and vector-search/RAG systems, I aim to build tools that are fast, measurable, and operationally useful.

Experience

Work history, roles, and key accomplishments

Current

Data Science Intern

Current

Acropolis Institute Of Bajaj Finserv Health Limited

Jul 2025 - Present (11 months)

Built a production cataract detection ML/DL system achieving 98.3% ROC-AUC and 94.8% accuracy using BioMedCLIP, Transformers, and MTCNN. Architected an IPD claims pipeline processing 1.2 lakh claims/month, reducing costs by INR 14L/month and manual effort by 61% while supporting INR 144 crores in claims.

Python SQL PyTorch Transformers Optuna BioMedCLIP OpenCV Langfuse fastAPI

Education

Degrees, certifications, and relevant coursework

Acropolis Institute

Bachelor of Technology (B.Tech), Computer Science

Grade: CGPA 7.45/10

Pursuing a B.Tech in Computer Science (CGPA: 7.45/10) with work on a production ML/DL cataract detection system. Reported performance includes ROC-AUC of 98.3% and 94.8% accuracy using BioMedCLIP and models including Transformers and MTCNN (tuned with Optuna).