Ambuj Agrawal — DevOps Engineer

Here’s a refreshed, human-sounding LinkedIn “About” you can paste (It is a meta joke :D) : I’m a Master’s student in Intelligent Information Systems at Carnegie Mellon (graduating Dec 2025), working at the intersection of NLP, evaluation, and scalable ML systems. This past summer I interned on the ML Engineering team at Netflix, where I fine-tuned a large speech model and built the retrieval + evaluation stack around it (automated metrics and a Labelbox human study). That work led to a clear lift in human preference and, more importantly, a faster, more trustworthy iteration loop for the team. Before CMU, I was a software engineer at LinkedIn and CRED, building zero-to-one, fault-tolerant microservices in Java/Go. I like the unglamorous parts too—clean interfaces, observability, and the debugging that makes systems reliable and cost-effective. I did my undergrad in IT at IIIT Allahabad, with research in NLP and human-robot interaction. I’m especially interested in applying language technologies to healthcare and consumer-facing products—making complex capabilities measurable, efficient, and useful in the real world. Outside of work, you’ll find me reading, playing badminton or playing drums. Open to conversations about ML engineering/research roles in evaluation, personalization, generative audio/agents, and healthcare AI.

Stackforce AI infers this person is a Machine Learning Engineer with a focus on NLP and scalable systems in the tech industry.

Location: Menlo Park, California, United States

Experience: 4 yrs 8 mos

Skills

Large Language Models (llm)
Mlops
Entrepreneurship
Text-to-speech Synthesis
Machine Learning
Go (programming Language)
Spring Framework
Java

Career Highlights

Fine-tuned speech models for Netflix, enhancing multilingual dubbing.
Built fault-tolerant microservices at LinkedIn and CRED.
Research in NLP and human-robot interaction at CMU.

Work Experience

Snowflake

Software Engineer - Machine Learning (5 mos)

Netflix

Machine Learning Intern (3 mos)

Carnegie Mellon University

Graduate Teaching Assistant (4 mos)

Machine Learning Research Assistant (1 yr 4 mos)

CRED

Senior Software Engineer (9 mos)

Software Engineer (7 mos)

Software Engineer (1 yr 7 mos)

CRED

Software Engineering Intern (4 mos)

Elucidata

Software Engineering Intern (3 mos)

Software Engineering Intern (2 mos)

Indian Institute of Technology (Banaras Hindu University), Varanasi

Machine Learning Research Intern (2 mos)

Education

Master's degree at Carnegie Mellon University

Bachelor of Technology at Indian Institute Of Information Technology Allahabad

Ambuj Agrawal

DevOps Engineer

Menlo Park, California, United States4 yrs 8 mos experience

AI EnabledAI ML Practitioner

Key Highlights

Fine-tuned speech models for Netflix, enhancing multilingual dubbing.
Built fault-tolerant microservices at LinkedIn and CRED.
Research in NLP and human-robot interaction at CMU.

Stackforce AI infers this person is a Machine Learning Engineer with a focus on NLP and scalable systems in the tech industry.

Contact

Skills

Core Skills

Large Language Models (llm)MlopsEntrepreneurshipText-to-speech SynthesisMachine LearningGo (programming Language)Spring FrameworkJava

Other Skills

Python (Programming Language)NLTKLarge Language Model Operations (LLMOps)Amazon Mechanical TurkMentoringDeep LearningPyTorchSpeech ProcessingTTSRetrieval-Augmented Generation (RAG)Information RetrievalNatural Language Processing (NLP)Apache PinotDatadogSpring Boot

About

Experience

4 yrs 8 mos

Total Experience

1 yr 5 mos

Average Tenure

5 mos

Current Experience

Snowflake

Software Engineer - Machine Learning

Jan 2026 – Present · 5 mos · Menlo Park, California, United States · On-site

Netflix

Machine Learning Intern

May 2025 – Aug 2025 · 3 mos · Los Gatos, California, United States · On-site

1. Built a full-catalog audio retrieval system at Netflix to choose reference clips for style/emotion transfer from English to target languages in synthetic dubbing; fine-tuned wav2vec2-2B with contrastive loss for multilingual dubbing embeddings.
2. Made an auto evaluation suite to automate the process of finding models/setups/hyperparameters that give better dub in the context of emotion/speaker style matching. Followed it up with a human evaluation study on Labelbox as well.
3. Evaluations show an improvement of ~7.5% with auto evals and ~13% win rate with human preference (CMOS: +0.2) in the synthetic voice. The finetuned embedding model outperformed all the off-the-shelf models.

Text-to-Speech SynthesisMachine LearningDeep LearningMLOpsPython (Programming Language)Large Language Models (LLM)+6

Carnegie mellon university

2 roles

Graduate Teaching Assistant

Jan 2025 – May 2025 · 4 mos · Pittsburgh, Pennsylvania, United States · On-site

Teaching assistant to the course: 67-428: Innovation & Entrepreneurship Course taught by Prof. Sara Moussawi

EntrepreneurshipMentoring

Machine Learning Research Assistant

Sep 2024 – Jan 2026 · 1 yr 4 mos · Pittsburgh, Pennsylvania, United States · On-site

Working as a Research Assistant under Dr. Fernando Diaz on the “Starlight” team, leveraging granular and implicit human preference feedback through infilling and edits to enhance personalization and reduce cognitive load in comparing LLM-based system outputs.

Python (Programming Language)NLTKLarge Language Models (LLM)Large Language Model Operations (LLMOps)MLOpsAmazon Mechanical Turk

Cred

2 roles

Senior Software Engineer

Promoted

Oct 2023 – Jul 2024 · 9 mos · Bengaluru, Karnataka, India · On-site

1. Built core components in the bill payments platform, powering services like electricity, rent, and gift vouchers.
2. Built zero-to-one products that added revenue stream for the company and handled spiky traffic of more than 100 QPS.

Apache PinotGo (Programming Language)DatadogSpring BootSpring FrameworkApache Kafka+1

Software Engineer

Feb 2023 – Sep 2023 · 7 mos · Bengaluru, Karnataka, India · On-site

1. Owned and built core backend services related to credit card management.
2. Solved a complex native memory leak in JVM due to memory fragmentation, cutting AWS cost by 25% for multiple services.
3. Developed a pipeline to deploy product config at CRED’s hackathon saving $150k annually; team placed third in 70+ submissions.

Go (Programming Language)DatadogDropwizardJavaSpring BootPython (Programming Language)+3

Software Engineer

Jul 2021 – Feb 2023 · 1 yr 7 mos · Bengaluru, Karnataka, India · On-site

Upgraded authentication cookie signatures to ECDSA-256 after testing on CPU, memory usage, and security benchmarks.
Fixed major vulnerabilities like replay attacks during Sign in with Google flow by caching server-side nonce.
Integrated sign-in with Facebook to move away from passwords with a 1% WAU increase and collaborated across verticals.
Optimized last seen at timestamp processing of users reducing resources by 50% by using Samza streams and efficient batching.

SQLNoSQLJavaSpring BootSpring FrameworkContinuous Integration (CI)

Cred

Software Engineering Intern

Feb 2021 – Jun 2021 · 4 mos · Bengaluru, Karnataka, India · On-site

1. Worked on making a complete backend microservice independently from scratch for a new line of utility on CRED to production with test coverage. Collaborated across verticals to help ensure the delivery of the product.
2. Worked on a microservice that handles 300M+ hits, enhancing and deprecating features.
3. Used Spring boot, dropwizard, AWS(SQS, DynamoDB, S3, Parameter Store, Lambda), Datadog, signalFX, SQL for development.

Spring Framework

Elucidata

Software Engineering Intern

Oct 2020 – Jan 2021 · 3 mos · Delhi, India

1. Worked with Pfizer as a client and developed different work‐flows for experiment log management that the researchers at Pfizer use to develop vaccines.
2. Worked in a small nimble team to continuously add features and timely fulfil user requirements in close collaboration with the Pfizer team. Extensively worked with Groovy on Grails and Oracle Relational Database(SQL).

JavaGrails

Software Engineering Intern

May 2020 – Jul 2020 · 2 mos

1. Worked and managed different states of phone numbers to delete the storage of old states when no longer required. Analyzed the nature of traffic of different types of handles(emails and phone number) using sensors.
2. Built various APIs for storing new way of phone numbers while deprecating the old way.
3. Worked on email and phone number verification system. Proposed an alternative to libphonenumber(google open-source library) for phone number verification.

JavaSpring Framework

Indian institute of technology (banaras hindu university), varanasi

Machine Learning Research Intern

May 2019 – Jul 2019 · 2 mos · Varanasi Area, India · On-site

Developed a robust high-accuracy model on a time series dataset (bpi-12) to improve Predictive Business Monitoring using LSTM and self-attention layer-based models.
Link: https://github.com/aka-jumba/predictive_process_management

Machine LearningPython (Programming Language)