Apoorv Reddy

Senior Software Engineer

Bengaluru, Karnataka, India12 yrs 9 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in Machine Learning and NLP.
Successfully deployed fraud detection model saving over $3M annually.
Led smart substitutions project featured in major publications.

Stackforce AI infers this person is a Machine Learning Engineer with expertise in AI and Fintech.

Contact

imp.apoorv@gmail.com LinkedIn

Skills

Core Skills

Machine LearningNatural Language Processing

Other Skills

AlgorithmsBashC++Data StructuresDeep Neural NetworksHTKJavaLaTeXLinuxMatlabOutlier Detection AlgorithmsPerlProgrammingPythonSignal Processing

Experience

Google

3 roles

Senior Software Engineer

Jun 2024 – Present · 1 yr 9 mos

LLM inference and training optimisations of Gemini Transformer models for search traffic.
Matformers for AI overviews. Post-training of matformed Gemini 2.0 Flash Lite for AI overviews and combining with routers for routing easy queries to nested/smaller models. Work done in collab with GDM
Explored param / inference cost space for nested models of 2.0 Flash-Lite and this work resulted in GDM making an informed decision on the architecture for nested models and then joint training.
RL training optimizations for AI mode. This involved running lots of ablation experiments on directions like quantized reward models, quantization aware training, importance sampling for off-policy RL. Also other small improvements like model sharding.

PythonMachine LearningSpeech TechnologySignal ProcessingNatural Language ProcessingAlgorithms

Senior Software Engineer

Promoted

Nov 2023 – May 2024 · 6 mos

Software Engineer Machine Learning

Dec 2021 – Nov 2023 · 1 yr 11 mos

Annotating knowledge graph entity mentions in queries.
These entity mentions help advertisers control branded traffic (See Brand Control feature in Google Ads).
Responsible for E2E quality of annotation systems.

Natural Language ProcessingData Structures

Walmart global tech

2 roles

Staff Data Scientist

Aug 2021 – Dec 2021 · 4 mos

Senior Data Scientist

May 2019 – Jul 2021 · 2 yrs 2 mos

Working on the smart substitutions problem at Walmart. My team's work has been featured on
1) TechCrunch: https://techcrunch-com.cdn.ampproject.org/c/s/techcrunch.com/2021/06/24/walmarts-ai-is-getting-smarter-about-grocery-delivery/amp/
2) Fast Company:
https://www.fastcompany.com/90555536/three-ways-walmart-has-created-a-better-store
3) Business Insider:
https://www.businessinsider.com/walmart-grocery-replacements

American express

Research Engineer

Nov 2016 – Apr 2019 · 2 yrs 5 mos · Bangalore

1) R&D for productionizing streaming machine learning algorithms for big data.
Successfully deployed AmEx's first streaming machine learning model for fraud detection in production, with an annual savings of >$3M per year.
Press coverage: https://www.forbes.com/sites/johnkoetsier/2020/09/21/50-less-fraud-how-amex-uses-ai-to-automate-8-billion-risk-decisions/
2) Deep neural networks (FFNNs, CNNs and AEs) for representation learning.
3) Outlier Detection Algorithms

Machine LearningDeep Neural NetworksOutlier Detection Algorithms

Cognitivescale

3 roles

Data Scientist

Apr 2015 – Oct 2016 · 1 yr 6 mos

All things IR - NLP (Query Understanding), Contextual Spell Check systems, Search Relevance, Text Classification, Business Domain Terms Extraction, Query Expansion,etc.

Natural Language ProcessingText Classification

Data Scientist

Dec 2014 – Mar 2015 · 3 mos

I'm working on search relevance here at CognitiveScale.

Natural Language Processing

Research Intern

Sep 2014 – Nov 2014 · 2 mos

1) Implemented a generic website scraping agent in Python based on HTML tagpath clustering.
2) Modeled the HTML of a webpage as a time series signal and extracted data from regions of interest.
3) Wrote generic Java code for importing any ontology into a Neo4j graph database. Used the OWL and Neo4j apis.
4) Descriptive keyphrase extraction and concept extraction from user reviews of travel attractions.

PythonJava

Interactive intelligence

Research Intern

Jan 2014 – May 2014 · 4 mos

Worked on developing solutions for spoken document retrieval (SDR) at Interactive Intelligence. Given a text query, goal was to retrieve all the spoken documents/files with high accuracy and as fast as possible. Our efforts resulted in a paper submitted to INTERSPEECH 2014.

Iiit hyderabad

MS by Research

May 2012 – Jun 2014 · 2 yrs 1 mo

I am working on spoken term detection for my MS thesis work. The work involves finding relevant features to index the speech database, and using these features to prune the search space for spoken term detection.

Speech Technology