Default profile banner
Sahil NayakSN

Sahil Nayak

@sahilk12nayak

Research Associate AI at Keywords Studios

Gurgaon, Haryana, India

Keywords StudiosIndian Institute of Technology, Roorkee

I am a Research Associate with a strong foundation in machine learning, deep learning, and generative AI. I am a graduate of IIT Roorkee, with a specialization in building scalable AI systems, including RAG-based chatbots and LLM evaluation pipelines. With hands-on experience in prompt engineering, NLP, and MLOps, I have improved model accuracy and performance across multiple projects and roles. I have worked on real-world applications involving large language models, quantitative finance, and data-driven decision-making, consistently delivering measurable improvements in efficiency and accuracy. I am passionate about leveraging AI to solve complex problems and building intelligent, production-ready systems.

Experience

Research Associate AI

Keywords Studios

Full-time•Jul 2025 - Present•Gurgaon, Haryana, India

Designed and evaluated LLM reasoning workflows using Step Level and advanced Chain of Thought (CoT) prompts to improve large language model reasoning, resulting in a 10% increase in overall performance accuracy. Applied python prompt engineering to automate text feature extraction and labeling, improving downstream agentic AI system accuracy by 12%. Built data-driven evaluation pipelines using pandas to analyze 1000+ LLM outputs, applying qualitative and quantitative metrics to improve response consistency and reduce failure cases by 15%.

AI prompt specialist

Outlier

Freelance•Oct 2024 - Mar 2025

Optimized generative AI models using PyTorch by performing STEM task training (physics/mathematics), image training, and adding cultural nuances through iterative feedback cycles. Designed and executed large-scale AI/ML data curation and evaluation pipelines, annotating and validating 500+ samples with scikit-learn statistical quality checks and KPI-based metrics, leading to a 40% improvement in downstream model accuracy and prediction consistency. Applied Python-based mathematical and statistical libraries to solve and validate 200+ complex tasks and containerized pipelines using Docker, enhancing model reasoning quality and overall task-level performance.

Education

Indian Institute of Technology, Roorkee

Bachelors of Technology

Chemical Engineering

Nov 2020 - Jul 2024•Grade: 6.56, First Class Honor

Licenses & Certifications

Google Cloud Arcade Facilitator Program

Google Cloud

Skills

Python
C++
SQL
Machine Learning
Deep Learning
Generative AI
Prompt Engineering
Natural Language Processing
GPT
BERT
RLHF
Probability
Statistics
Feature Engineering
Model Evaluation
Statistical Analysis
Hypothesis Testing
Probability & Distributions
A/B Testing
CI/CD
Scikit-Learn
TensorFlow
Numpy
Pandas
Transformer
PyTorch
Matplotlib
Seaborn
Plotly
Cufflinks
Looker
Git
GitHub
PostgreSQL
GCP
FastAPI
API
Docker
Kubernetes
RAG
Vector Databases
LLM Evaluation
LangChain