Manish V.

AI Researcher

San Francisco, California, United States10 yrs 8 mos experience

Highly Stable

Key Highlights

Expert in Machine Learning and Deep Learning applications.
Proven track record in building scalable ML systems.
Strong background in data science and statistical modeling.

Stackforce AI infers this person is a Machine Learning Engineer with expertise in B2C applications and deep learning technologies.

Contact

Skills

Core Skills

Machine LearningDeep LearningDistributed SystemsData Science

Other Skills

Programming LanguagesPyTorchHadoopStatisticsPandas (Software)Deep Neural Networks (DNN)KerasTensorBoardStatistical ModelingGoogle Cloud Platform (GCP)TensorFlowRecurrent Neural Networks (RNN)ScalaMapReduceComputer Vision

About

Passionate Machine Learning Researcher and Software Developer leading large scale deep learning applications. Team player with a can-do attitude, phenomenal time management skills, and a strong user focus. Has developed several full stack web apps and distributed systems. Vibrant personality awarded for Innovation and diligence. GitHub - https://github.com/manish181192; OpenAi -https://gym.openai.com/users/manish181192

Experience

10 yrs 8 mos

Total Experience

2 yrs

Average Tenure

8 mos

Current Experience

Netflix

Machine Learning Engineer

Aug 2025 – Present · 8 mos · Los Angeles Metropolitan Area · Remote

Improving Ads Members experience and building ML systems for NetflixAds

Machine LearningDeep Learning

Snap inc.

Machine Learning Engineer

May 2023 – Aug 2025 · 2 yrs 3 mos

Ad Snapchat user Identity graph and measurability.
Build Distributed ML training framework for Attribution and Measurement

Machine LearningDistributed Systems

Twitter

2 roles

Senior Data Scientist

Promoted

May 2021 – Mar 2023 · 1 yr 10 mos

Engagement ranking Models calibration and weight tuning
Home timeline ranking model is a weighted mixture of engagement objectives, the weights dictate the behavior of the tweets we show to the users.
Calibration of the model refers to correcting the errors in the model predictions based on the observed probability of engagement.
We further tune the weight of the models by creating an A/B test exploration framework using Hill climbing approach to maximize our business objectives
Offline candidate generation simulations and causal evaluation
Adding features to candidate generation models and evaluation often takes more than a month of time due to A/B testing and it's often hard to predict possible gains for the ranking models.
Developed the first ever off-prem candidate generation algorithm simulation that generated tweets for users home timeline using scalable embedding ANN search.
Developed the offline causal evaluation of counterfactual tweet candidates using new signals by estimating the marginal gains in Ranking model scores.

Programming LanguagesPyTorchHadoopStatisticsPandas (Software)Deep Neural Networks (DNN)+8

Machine Learning Engineer II

Feb 2020 – May 2021 · 1 yr 3 mos

Notification Personalization
Every user has a different notification consumption behavior. I personalize the user’s notification volume and time schedule based on their consumption pattern using a two tower neural network based counterfactual prediction model that maximizes our business objectives.
Patent pending Twitter Ref: TP12233US01: Notification Volumes in Messaging platform

Machine LearningDeep Learning

Latent ai, inc.

Machine Learning Intern

Sep 2019 – Dec 2019 · 3 mos · New Jersey

Deep learning model compression using distillation based quantization of speech detection algorithms for 8 bit FP precision audio headphone chips.
Mini Alexa without internet for headphones

Machine LearningDeep Learning