Nikhil Kumar
Experienced ML Engineer and NLP Developer specialized in Medical Coding, Entity Disambiguation, and information extraction from financial documents.
- Role
ML & NLP Engineer
- Years of Experience
3.67 years
- Professional Portfolio
View here
Skillsets
- Ift
- VectorDBs
- Transformers
- Retrieval Augmented Generation
- QLoRA
- Machine Learning
- LoRA
- LLama2
- Amazon SageMaker
- Generative AI
- Deep Learning
- LLMs
- PyTorch
- NLP
- SQL
- Python
- Python
- Sft
- Prompt Engineering
- vertexAI
- MLFlow
- Generative AI
- BERT
- LLMs
- MLOps
- FastAPI
- PyTorch
- NLP
- Docker
- Git
- GCP
- SQL
Professional Summary
3.67Years
- Jan, 2026 - Present 4 months
Senior Data Scientist
Target - Aug, 2024 - Jan, 20261 yr 5 months
Data Scientist
Target - Feb, 2023 - Jul, 20241 yr 5 months
Machine Learning Engineer
Arintra - Apr, 2022 - Nov, 2022 7 months
NLP Developer
Fluid AI
Applications & Tools Known
Python
SQL
.png)
Docker
Git
.png)
FastAPI
Machine Learning
Deep Learning
NLP
MLOPS
MLFlow
GCP
Amazon SageMaker
VertexAI
Generative AI
LLMs
Prompt Engineering
QLoRA
Work History
3.67Years
Senior Data Scientist
TargetJan, 2026 - Present 4 months
Data Scientist
TargetAug, 2024 - Jan, 20261 yr 5 months
Machine Learning Engineer
ArintraFeb, 2023 - Jul, 20241 yr 5 months
Built and fine-tuned NLP models for medical coding, entity disambiguation, relation extraction, and question-answering. Improved various metrics and deployed scalable inference engines.
NLP Developer
Fluid AIApr, 2022 - Nov, 2022 7 months
Achievements
- Mention Extraction for Medical Coding with F1 score of 88%
- Reduced the model inference latency by 30% by converting to ONNX format
- Performed SFT on llama2-13bmodel with 87% accuracy
- Used SAPBert for entity disambiguation and reduced latency by 40%
- Built a relation extraction model with F1 score of 77%
- Achieved 91% F1 score in Question-Answering
- Led data annotation team and designed annotation processes
Major Projects
3Projects
Mention Extraction for Medical Coding
Built a NER model to extract mentions in patient charts for medical coding using UmlsBERT. Achieved an F1 score of 88% and reduced inference latency.
Efficient Entity Disambiguation
Used SAPBert with augmented UMLS for entity disambiguation and deployed using VertexAI. Reduced manual database additions and latency.
Delinquency Prediction
Built prediction models for Latin American banks using XGBoost. Achieved 72% recall and reduced financial risks.
Education
Bachelor of Engineering (B.E) in Electronics and Communication
R.V. College of Engineering (2017)Masters in Artificial Intelligence and Machine Learning
BITS Pilani (2024)
Certifications
Nlp specialization