Hi, I am

Govind Thakur. I build AI systems that improve lives.

AI Systems Engineer specializing in production-grade LLM and multimodal systems. I build large-scale AI-enabled APIs serving 1M+ users, design distributed data pipelines, and develop evaluation platforms that translate experimentation into product decisions.

About & Education

Background

I am an AI Systems Engineer with a deep focus on bridging the gap between cutting-edge model experimentation and reliable, large-scale production deployments.

Currently at American Express, I drive multiple high-impact AI initiatives. Beyond building GenAI enterprise Search APIs serving 1M+ users and architecting the AGX evaluation platform, I am part of a specialized team developing an intelligent GenAI voice bots initiative designed to transform and elevate customer service experiences.

My expertise covers the full MLOps lifecycle: from deploying distributed data pipelines via Airflow and Spark, to designing production-grade LLM evaluation frameworks and semantic search optimization.

Education

Master of Science, Computer Science – Artificial Intelligence

University of Southern California (USC)
Aug 2022 – May 2024

Bachelor of Engineering, Computer Engineering

Dwarkadas J Sanghvi College of Engineering
Jul 2018 – Jun 2022

Technical Skills

Core Competencies

AI Engineering
ML Systems
LLMs
Semantic Search
Generative AI
RAG
Deep Learning

Tech Stack

Python
C++
Java
SQL
React
FastAPI
Flask
TensorFlow
PyTorch
Docker
AWS (SageMaker, EC2, S3)

Tools & Platforms

GitHub Actions
Jenkins
Kubernetes
SonarQube
Splunk
Grafana
CI/CD
Cloud Deployments

Professional Experience

AIML Engineer-II, GenAI Search Team

American Express
📍 Phoenix, AZ 🗓 Sept 2024 – Present
  • Designed and maintained production Search APIs (Python, Java) powering GenAI Search for 1M+ monthly mobile users; optimized retrieval pipelines and parallelization to reduce latency across releases.
  • Led end-to-end development of AGX, a full-stack (React + Python) internal search evaluation platform enabling engineers to simulate production behavior, analyze quality metrics, and detect regressions prior to release.
  • Defined and operationalized search quality metrics (NDCG@k, precision/recall, containment, latency) across 1M+ monthly user traffic to guide production release decisions.
  • Led LLM-based experimentation initiatives (RAG, ranking refinement, containment optimization), delivering AGX-powered executive demos that translated large-scale metric analysis into product and governance decisions.
  • Architected automated regression and content validation pipelines (Python, GitHub Actions) running daily to detect application runtime failures and search content defects, reducing regression defects by 35% and removing significant manual QA effort.

Machine Learning Engineer

Northern Lights
📍 Los Angeles, CA 🗓 Jun 2023 – Sept 2024
  • Co-led architecture and implementation of a multimodal analytics MVP, designing distributed ingestion pipelines and backend integrations for production deployment.
  • Built containerized ML workflows on AWS SageMaker and Docker for automated training, evaluation, and CI/CD delivery.
  • Designed Python Airflow DAGs orchestrating data pipelines processing millions of records for multimodal model training.
  • Deployed and productionized LLM-based systems to extract structured insights from unstructured enterprise data.

Software Engineering Intern

JP Morgan Chase & Co
📍 Mumbai, India 🗓 Jun 2021 – Aug 2021
  • Improved NLP accuracy of JPMC's E-Trading Assistant Bot by 20% through Rasa model tuning and advanced intent classification.

Patents & Projects

Patent Pending

System and Method for Estimating Intrinsic Popularity of Content

US Serial No. 63/642,980. Developed predictive models for ranking content popularity, improving engagement metrics in adaptive systems.

Predictive Modeling Ranking Systems
Patent Registered

Automated Attendance Management System

ROC No. L-110318/2022. Designed an automated ecosystem to efficiently manage user attendance without relying on manual intervention.

Automation Systems Architecture
Project • Dec 2020 – Jun 2022

Speech to Code: Voice-Driven Developer Environment

Built a PyTorch-based voice-controlled coding interface using LSTM seq2seq targeting visually impaired developers. Integrated StackOverflow API for autonomous error correction and contextual code suggestions.

PyTorch LSTM seq2seq API Integration
Project

TextGuard: NLP Toxicity Detection

Multi-model NLP system using BERT masked language modeling to detect and reverse toxic content. Achieved 87% toxicity reduction maintaining text coherence.

BERT NLP Python
Project

Sentiment Analysis API

Production-ready sentiment analysis system deployed on AWS SageMaker. Built end-to-end ML pipeline representing a complete MLOps workflow.

AWS SageMaker Deep Learning REST API
Project

Plagiarism Detection System

NLP-based plagiarism detection deploying text similarity algorithms. Deployed as production-ready REST API on AWS SageMaker with 92% detection accuracy.

Python NLP AWS SageMaker

Research, Certifications & Honors

Publications & Research

Emergent Ethics in Agentic Simulations: Moral Incubation and Conscious Development across GPT, Claude, Gemini, and XAI

Springer Nature – AI and Ethics Journal (Under Peer Review, 2025)

Thakur, G.

Comparison of Tabular Synthetic Data Generation Techniques using Propensity and Cluster Log Metric

Elsevier – International Journal of Information Management Data Insights (2023)

Pathare, A., Mangrulkar, R., Thakur, G., et al.

Google Scholar Profile

Certifications & Honors

  • IBM AIML Specialization
  • Udacity ML Nanodegrees (3)
  • AWS Certified Cloud Practitioner (Planned)

Extracurriculars

  • Course Producer for USC CSCI 401 & 102
  • Mentored 50+ students in ML and GenAI projects, providing architectural guidance and debugging distributed systems issues.

Restricted Access

Gaia is an AI assistant specifically trained to discuss my professional background. Please enter the recruiter access code.

Invalid Access Code
Gaia AI Assistant