Vinit Sutar

Nov 2025 — Present

A5E Consulting (Client: Nielsen Media) | Bangalore

Architected end-to-end LLM system computing Pre-Game Watchability Index across 14 geographies for Amazon Prime Video Marketing ad-placement targeting.
Engineered Strands agent data ingestion pipeline aggregating ELO scores, team form, head-to-head metrics, and 5-year cross-platform fan sentiment.
Reduced model complexity by consolidating to single Claude Sonnet model with custom deterministic scoring rubric, validated via LLM-as-a-Judge pipeline.
Directed AI-assisted frontend development using KIRO agent for real-time LLM observability dashboard on serverless AWS stack (S3 + CloudFront).
Integrated Claude Code with custom skill sets via Repowise to reduce per-task token costs at scale.

Mar 2025 — Nov 2025

Shriram Finance Limited | Bangalore

Led and mentored a team of 5 Data Scientists, upskilling them in GenAI architectures and advanced ML techniques.
Improved Propensity Models increasing conversions across Fixed Deposits products by INR 520 Crores.
Implemented Market Mix Modelling best practices, optimizing marketing campaigns to save 5% on spends.
Engineered scalable RAG architecture on Azure Databricks with HNSW vector indexing and Tool Calling with Gemma 3.

Apr 2024 — Feb 2025

Conneqt Digital (Client: Anheuser-Busch InBev) | Bangalore

Developed GLMNET and Logistic Regression attribution models with 98% accuracy and conversion package with ±5% error.
Implemented 2^(n-1) fractional factorial designs and CUPED-backed Total Distribution approaches, halving RCT time.
Built propensity score matching pipeline (DTW, LSH, Faiss) for 200k stores to optimize Control/Treatment grouping.

July 2022 — Apr 2024

Impact Analytics | Bangalore

Modernised multi-hierarchy Prophet pipeline, increasing store allocation regression accuracy by 3%.
Designed DTW-based store clustering improving lost-sales imputation from 0.3% to 0.7%.
Mentored team of two on inventory pipelines and client knowledge transfer.

Apr 2021 — July 2022

HiMedia Laboratories | Mumbai

Improved micro-organism detection accuracy by 40% using OpenCV compared to existing algorithm.
Developed annotation pipeline using Raspberry Pi saving INR 80/image for 500 images across 15 species.
Implemented Active Learning using Mask RCNN saving INR 6,00,000 in annotation costs.
Saved INR 1,00,000 in instrument costs while increasing accuracy from 80% to 94%.

Professional Experience