Sahil Kumar

AI Researcher

Gurugram, Haryana, India2 yrs 2 mos experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

Achieved 88% accuracy in NLP model improvements.
Developed advanced video classification algorithms.
Strong blend of business and technology expertise.

Stackforce AI infers this person is a Data Science professional with a focus on NLP and Machine Learning in SaaS.

Contact

Skills

Core Skills

Natural Language Processing (nlp)Machine LearningData Analysis

Other Skills

Analytical SkillsArtificial Intelligence (AI)Big QueryCommunicationData AnalyticsData CleaningData ManagementData ModelingData ModelsData ScienceData VisualizationDeep LearningExploratory Data AnalysisFastAPIGitHub

About

An Emerging Data Scientist/ Data Analyst armed with a Certification in Data Science, highly motivated and results-driven professional with an MCA in Data Science and Machine Learning. Graduation in commerce and master's degree in computer science provide a strong mix of expertise in both domains as business management and technology, allowing to think from both ends and provide effective solutions for business problems.

Experience

2 yrs 2 mos

Total Experience

5 mos

Average Tenure

11 mos

Current Experience

India today

AI Engineer

Jul 2025 – Present · 11 mos · Noida, Uttar Pradesh, India · On-site

Collegedunia

Data Scientist

Jun 2024 – Sep 2024 · 3 mos · Gurugram, Haryana, India · On-site

Build News Summary Algorithm
1. Improved the performance of the "facebook_bart_large_cnn" summarization model by increasing accuracy from 60% to 88% using Gemma2-2b-it.
2. Condensed articles into 85-word summaries using LLMs and advanced prompting techniques like Chain of Thought and Self-Consistency.
3. Solved text length limitations by integrating LangChain for handling long documents and deployed the tool using FastAPI.
Video Classification Algorithm
1.Build a video classification model to categorize video into 10 groups using the supervised learning techniques. Focus on English or non-audio content.
2.Filter the non-english content using OpenAI model Whisper and also generate audio transcript. Extract the frames, convert them image to text using MoonDream2 Model.
3.Combined the text outputs with metadata of video and applied Llama model with Chain of Thoughts and Self Consistency prompting to refine classification metrics.
4.Achieved 85% classification accuracy by leveraging advanced prompting techniques and metadata analysis.