Vishesh Kaul profile photo 1 on Tal

Vishesh Kaul

Lead DS @ Cocoblu | IISc Bangalore | NIT Hamirpur

Greater Bengaluru Area

1+ year of work experience

Cocoblu Retail
Jul 20264 mosPresent

Built a 9-model forecasting tournament (SARIMAX, LightGBM, TimesFM, Mamba) with automated backtesting to select a champion model per line. • Architected a GCP-native MDM (Master Data Management) Portal (Cloud Run, Secret Manager, IAM RBAC) for self-service data access across business verticals. Built fault-tolerant GCS→BigQuery pipelines (Cloud Functions, Pub/Sub) with automated schema-drift detection. Automated BigQuery-to-Excel reporting and shipped Looker Studio dashboards for merchandising and ops.
Cocoblu Retail
Jan 20267 mosJul 2026

Engineered an MoE-based AI cataloging system using AWS Bedrock and FastAPI, reducing processing time by 50% while enforcing strict schema compliance. ​Developed a proprietary hallucination detection framework utilizing Geometric Algebra (\wedge) to capture bivector embedding drift and flagging factual anomalies with high precision via Cross-Encoder softmax weights. ​Architected a large-scale market research module benchmarking 100K+ products using FashionCLIP embeddings, driving a 50% reduction in operational resource effort. ​Built a hybrid demand forecasting pipeline by ensembling time-series foundation models with classical frameworks (SARIMAX/ARIMA), achieving a 5% MAPE for short-horizon predictions.
iMerit Technology
May 20253 mosJul 2025

Bengaluru, Karnataka, India

Optimized AWS S3 data pipelines using Glue Jobs and Crawlers, improving file compaction and query efficiency. Built real-time analytics dashboards with AWS QuickSight & Athena, enabling client-ready insights. Developed custom inference logic for MBPTrack3D, processing ground truth JSON and 3D Point Cloud sequences to generate bounding box outputs across frames. Containerized the MBPTrack3D pipeline with Docker and integrated FastAPI, making the 3D object tracking system server-agnostic and easy to deploy. Improved tracking accuracy by 2.5% for abrupt movements by implementing BOTSort-inspired IoU gating and appearance-based association for unmatched detections.
Indian Institute of Technology, Kanpur
Jun 20232 mosJul 2023

Kanpur, Uttar Pradesh, India

Worked under Prof Avinash Kumar Agarwal at Engine Research Laboratory
Tenneco
Jun 20222 mosJul 2022

Parwanoo

Vishesh’s AI Stack

anthropic apis
anthropic apis
antigravity
antigravity
claude code
claude code
claude
claude
gemini
gemini
hugging face
hugging face

Education

Indian Institute of Science (IISc)

Master of Management, Data Science & Business Analytics

Indian Institute of Science (IISc)

20242 yrs2026
National Institute of Technology Hamirpur-Alumni

Bachelor of Technology - BTech, Mechanical Engineering

National Institute of Technology Hamirpur-Alumni

20204 yrs2024

Join Vishesh and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Vishesh Kaul on Tal

Headline

Lead DS @ Cocoblu | IISc Bangalore | NIT Hamirpur

About

Understand, then execute.

Location

Greater Bengaluru Area

AI stack

anthropic apis

antigravity

claude code

claude

gemini

hugging face

Work experience

Lead Engineer - Data Science at Cocoblu Retail, Jul 2026 - Present · 4 mos, Built a 9-model forecasting tournament (SARIMAX, LightGBM, TimesFM, Mamba) with automated backtesting to select a champion model per line. • Architected a GCP-native MDM (Master Data Management) Portal (Cloud Run, Secret Manager, IAM RBAC) for self-service data access across business verticals. Built fault-tolerant GCS→BigQuery pipelines (Cloud Functions, Pub/Sub) with automated schema-drift detection. Automated BigQuery-to-Excel reporting and shipped Looker Studio dashboards for merchandising and ops.

Data Science Intern at Cocoblu Retail, Jan 2026 - Jul 2026 · 7 mos, Engineered an MoE-based AI cataloging system using AWS Bedrock and FastAPI, reducing processing time by 50% while enforcing strict schema compliance. ​Developed a proprietary hallucination detection framework utilizing Geometric Algebra (\wedge) to capture bivector embedding drift and flagging factual anomalies with high precision via Cross-Encoder softmax weights. ​Architected a large-scale market research module benchmarking 100K+ products using FashionCLIP embeddings, driving a 50% reduction in operational resource effort. ​Built a hybrid demand forecasting pipeline by ensembling time-series foundation models with classical frameworks (SARIMAX/ARIMA), achieving a 5% MAPE for short-horizon predictions.

Data Engineering Intern at iMerit Technology, May 2025 - Jul 2025 · 3 mos, Bengaluru, Karnataka, India, Optimized AWS S3 data pipelines using Glue Jobs and Crawlers, improving file compaction and query efficiency. Built real-time analytics dashboards with AWS QuickSight & Athena, enabling client-ready insights. Developed custom inference logic for MBPTrack3D, processing ground truth JSON and 3D Point Cloud sequences to generate bounding box outputs across frames. Containerized the MBPTrack3D pipeline with Docker and integrated FastAPI, making the 3D object tracking system server-agnostic and easy to deploy. Improved tracking accuracy by 2.5% for abrupt movements by implementing BOTSort-inspired IoU gating and appearance-based association for unmatched detections.

Research Intern at Indian Institute of Technology, Kanpur, Jun 2023 - Jul 2023 · 2 mos, Kanpur, Uttar Pradesh, India, Worked under Prof Avinash Kumar Agarwal at Engine Research Laboratory

Internship Trainee at Tenneco, Jun 2022 - Jul 2022 · 2 mos, Parwanoo

Education

Master of Management, Data Science & Business Analytics, Indian Institute of Science (IISc)

Bachelor of Technology - BTech, Mechanical Engineering, National Institute of Technology Hamirpur-Alumni