Lalit Upadhyaya profile photo 1 on Tal

Lalit Upadhyaya

Senior AI Engineer @ Best Buy™ India | Deploying AI solution for Ads

Greater Bengaluru Area

9+ years of work experience

Best Buy™ India
Apr 20251 yr 6 mosPresent

Bengaluru, Karnataka, India

1. Deployed CTR and CVR prediction services on Kubernetes, handling 400+ RPS at sub-50 ms p95 latency. 2. Built a sequential session encoder that turns user click history into real-time embeddings, using cosine similarity against product vectors to feed ranking models. 3. Set up tracing and metrics using Cloud Trace Explorer, Prometheus, and Grafana to track RPS, p95/p99 latency, and error rates while cutting down log volume. 4. Developed agentic framework using golang ADK to parse production logs and pinpoint errors fast during peak holiday traffic.
Chubb
Nov 20231 yr 7 mosMay 2025

Bengaluru, Karnataka, India

1. Built an LLM-as-a-Judge pipeline to automatically test model outputs for accuracy, relevance, and hallucinations, cutting release cycles from 12 weeks to 6 weeks. 2. Reduced model inference latency from 800 ms to 300 ms (60% drop) by converting models to ONNX, adding in-memory caching, and using async calls. 3. Added SonarQube checks, automated unit/regression tests, and Locust load testing into the CI/CD pipeline to catch bugs and performance drops before release.
KPMG India
Jan 20221 yr 11 mosNov 2023

Bangalore Urban, Karnataka, India

1. Migrated 30+ analytics and reporting pipelines from AWS to GCP BigQuery, switching from ETL to ELT across batch, streaming, and SFTP workloads. 2. Built a streaming pipeline for GPS data using REST APIs, Cloud Functions, and Pub/Sub, dropping end-to-end latency from 300 ms to 20 ms. 3. Cut BigQuery query costs by 80% using partition pruning, incremental loads, and materialized views to support fast data access for models and dashboards. 4. Automated pipeline runs and monitoring using CI/CD and Airflow to keep data feeds reliable and prevent pipeline failures.
Tata Consultancy Services
Dec 20172 yrs 8 mosJul 2020

Greater Bengaluru Area

Built and scaled data and NLP engineering solutions supporting enterprise catalog discovery and analytics. Architected an end-to-end OCR, document ingestion, and translation pipeline to extract unstructured product metadata from catalog images and translate content across 7 South Asian languages. Implemented vector-based semantic search workflows using Python NLP libraries to optimize information retrieval, while designing and maintaining scalable Python and SQL data pipelines that powered mission-critical KPI dashboards and business reporting.
Tata Consultancy Services
May 20161 yr 7 mosNov 2017

Bengaluru, Karnataka

Collaborated closely with cross-functional teams and enterprise clients to translate ambiguous business requirements into robust, automated data systems, establishing reliable operational benchmarks and high data quality standards across distributed systems.
Tata Consultancy Services
Dec 20155 mosApr 2016

Gujarat, India

Trained on various technologies to onboard a project.

Lalit’s AI Stack

gemini
gemini
langchain
langchain
langgraph
langgraph
Google ADK
Google ADK
PydanticAI
PydanticAI
vLLM
vLLM

Education

Dhirubhai Ambani University

Msc in Data Science

Dhirubhai Ambani University

20202 yrs2022
Jorhat Engineering College

Engineer’s Degree, Electrical Engineering

Jorhat Engineering College

20114 yrs2015

Join Lalit and 10k+ other professionals on

Tal Boss

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

Download Tal Boss

Lalit Upadhyaya on Tal

Headline

Senior AI Engineer @ Best Buy™ India | Deploying AI solution for Ads

About

As a Senior AI Engineer at Best Buy™ India, I specialize in deploying real-time machine learning systems and AI solutions to optimize ad performance. With over 7 years of experience in data engineering and MLOps, I focus on building resilient AI infrastructure that supports high-throughput applications, ensuring low latency and operational reliability. My work includes deploying CTR and CVR prediction services, designing session encoding models, and implementing robust monitoring solutions with Prometheus and Grafana. My mission is to bridge the gap between cutting-edge machine learning research and practical, production-ready systems. By leveraging tools like Kubernetes, Cloud Trace Explorer, and ADK, I contribute to creating scalable and efficient platforms that drive measurable business outcomes. Passionate about real-time data processing and system optimization, I strive to enable seamless AI adoption for impactful decision-making.

Location

Greater Bengaluru Area

AI stack

gemini

langchain

langgraph

Google ADK

PydanticAI

vLLM

Work experience

Senior AI Engineer at Best Buy™ India, Apr 2025 - Present · 1 yr 6 mos, Bengaluru, Karnataka, India, 1. Deployed CTR and CVR prediction services on Kubernetes, handling 400+ RPS at sub-50 ms p95 latency. 2. Built a sequential session encoder that turns user click history into real-time embeddings, using cosine similarity against product vectors to feed ranking models. 3. Set up tracing and metrics using Cloud Trace Explorer, Prometheus, and Grafana to track RPS, p95/p99 latency, and error rates while cutting down log volume. 4. Developed agentic framework using golang ADK to parse production logs and pinpoint errors fast during peak holiday traffic.

Senior Data and Machine Learning Operations Engineer at Chubb, Nov 2023 - May 2025 · 1 yr 7 mos, Bengaluru, Karnataka, India, 1. Built an LLM-as-a-Judge pipeline to automatically test model outputs for accuracy, relevance, and hallucinations, cutting release cycles from 12 weeks to 6 weeks. 2. Reduced model inference latency from 800 ms to 300 ms (60% drop) by converting models to ONNX, adding in-memory caching, and using async calls. 3. Added SonarQube checks, automated unit/regression tests, and Locust load testing into the CI/CD pipeline to catch bugs and performance drops before release.

Cloud Consultant at KPMG India, Jan 2022 - Nov 2023 · 1 yr 11 mos, Bangalore Urban, Karnataka, India, 1. Migrated 30+ analytics and reporting pipelines from AWS to GCP BigQuery, switching from ETL to ELT across batch, streaming, and SFTP workloads. 2. Built a streaming pipeline for GPS data using REST APIs, Cloud Functions, and Pub/Sub, dropping end-to-end latency from 300 ms to 20 ms. 3. Cut BigQuery query costs by 80% using partition pruning, incremental loads, and materialized views to support fast data access for models and dashboards. 4. Automated pipeline runs and monitoring using CI/CD and Airflow to keep data feeds reliable and prevent pipeline failures.

Analyst at Tata Consultancy Services, Dec 2017 - Jul 2020 · 2 yrs 8 mos, Greater Bengaluru Area, Built and scaled data and NLP engineering solutions supporting enterprise catalog discovery and analytics. Architected an end-to-end OCR, document ingestion, and translation pipeline to extract unstructured product metadata from catalog images and translate content across 7 South Asian languages. Implemented vector-based semantic search workflows using Python NLP libraries to optimize information retrieval, while designing and maintaining scalable Python and SQL data pipelines that powered mission-critical KPI dashboards and business reporting.

Junior Analyst at Tata Consultancy Services, May 2016 - Nov 2017 · 1 yr 7 mos, Bengaluru, Karnataka, Collaborated closely with cross-functional teams and enterprise clients to translate ambiguous business requirements into robust, automated data systems, establishing reliable operational benchmarks and high data quality standards across distributed systems.

Initial Learning Programme at Tata Consultancy Services, Dec 2015 - Apr 2016 · 5 mos, Gujarat, India, Trained on various technologies to onboard a project.

Education

Msc in Data Science, Dhirubhai Ambani University

Engineer’s Degree, Electrical Engineering, Jorhat Engineering College