Nikhil Singh
Data Scientist at Tealbox Digital | Indian Institute of Technology, Kanpur
- Role
- Data Scientist at Tealbox Digital
- Location
- Gurugram, HR, IN
- LinkedIn followers
- 500 followers
About Nikhil Singh
Hey! I\'m an IIT Kanpur graduate and a Versatile professional with 1+ years of…
Experience
Data Scientist
Aug 2024 — Present · Indore, IN
Project 1: Creating an Audience Recommendation System using the Two-Tower Model (*Achieved 30% ROI based on the recommendations*)• Developed a Two-Tower Recommendation Model to provide audience recommendations for the ads over the Tik Tok & Meta platforms• Developed end-to-end pipelines to preprocess BigQuery data, generate multimodal embeddings for creative assets using ImageBind,and create audience embeddings with BERT, assigning the KPI specific performance labels, and preparing datasets for model training.• Created separate creative and audience towers to compute vector similarity between media and audience data, using KPI-based labels.• Passed the embeddings through the trained creative and audience towers and transformed them into the same vector space.• Used the Qdrant vector database to store the transformed embeddings and then later perform nearest neighbor search using SCANN.• Optimized model performance using AdamW optimizers and doing hyperparameter tuning to improve predictive accuracy up to 80.6%.Project 2:Conversational AI Chat System using RAG & Multi-Agent Supervisor• Built a chatbot using RAG architecture with LangChain and Gemini LLM, integrating multiple specialized agents managed by a supervisor agent.• Designed pipelines to process CSV performance data, generate vector embeddings with HuggingFace, and store/query them using Qdrant.• Developed an agent graph (creative strategist, media buyer, analyst agents) with capabilities like Python code execution, graph generation, and response streaming.• Fine-tuned supervisor prompts for optimal task delegation and conversion of agent outputs into actionable insights.Project 3: Created Multiple AI Agents for Multimodal Analysis and Onboarding Clients• Built AI agents using Gemini LLMs + LangChain, fine-tuned with task-specific prompts for creative analysis and client onboarding.• Developed agents to auto-generate ad descriptions and tags by analyzing media assets,identifying element
Education
Indian Institute of Technology, Kanpur
Bachelor of Technology - BTech
2020 — 2024
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.