Staff ML Engineer @ Adobe | LLMs, VLMs, Agents, Post-Training & Multimodal Retrieval
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Texas A&M University
Master’s Degree, Machine Learning/ Data Science
January 1, 2016 – January 1, 2018
RV College Of Engineering
Bachelor’s Degree, Electrical and Electronics Engineering
January 1, 2012 – January 1, 2016
Adobe
Staff Machine Learning Engineer
January 1, 2026 – Present
San Francisco Bay Area
Adobe
Senior Machine Learning Engineer
January 1, 2024 – January 1, 2026
San Francisco Bay Area
Adobe
Machine Learning Engineer 3
January 1, 2021 – January 1, 2024
San Francisco Bay Area
Adobe
Search and Machine Learning Engineer
August 1, 2019 – January 1, 2021
San Francisco Bay Area
Siemens
Deep Learning Intern
September 1, 2018 – December 1, 2018
Princeton, New Jersey
Texas A&M University
Graduate Assistant
January 1, 2018 – May 1, 2018
Bryan/College Station, Texas Area
Siemens
Deep Learning Intern-NLP and Computer Vision
June 1, 2017 – December 1, 2017
Greater New York City Area
Pervazive Inc.
Internship
August 1, 2015 – February 1, 2016
Bengaluru Area, India
Indian Institute of Science
Summer Internship
May 1, 2014 – August 1, 2014
Bengaluru Area, India
Deblurring Network Fine-Tuning Using Cycle-Consistent Adversarial Networks
January 1, 2018 – May 1, 2018
We propose an end-to-end trainable general framework to fine-tune any deblur- ring network. The model is a novel cyclically trained frame-work for fine-tuning pre-trained deblurring networks using a self-supervised approach. We fine-tune two trained DeblurGANs (a conditional GAN trained on Wasserstein loss) connected in series, one to deblur a blurry image and another to blur the deblurred image using opposite objectives. This yielded marginal improvement in the perceptual quality of the deblurred images.
Turney Algorithm for Sentiment Classification
April 1, 2017 – Present
Turney algorithm is an unsupervised sentiment classification algorithm based on the Pointwise Mutual Information (PMI) of the phrasal lexicon. The implementation was carried out on the IMDB movie review dataset in Python. Certain phrases were selected from the document based on their parts of speech tags. For each phrase, a PMI was calculated based on its proximity to the words "great" and "poor". For each review, the PMI of all phrases in that review are added and the polarity of the sum decides the class of the review. A 10-fold cross validation was carried out to average the results
Breast Cancer Survival Prediction using Gene Expression Data
April 1, 2017 – Present
The objective of this project was to predict if a breast cancer patient would survive looking at the patient's gene expression data. The top features (genes) were selected with feature selection algorithms like Sequential Forward Search (greedy algorithm) and Exhaustive Search using a wrapper approach. Secondly, simple algorithms like 3-Nearest Neighbours and Diagonal Linear Discriminant Analysis were designed on the selected features and proved to be effective as the dataset was small. The project was implemented in the R programming language.
Sentimemt Analysis with Naive Bayes and Peceptron Algorithm
March 1, 2017 – Present
A multinomial and a binarized Naive Bayes classifier was implemented for the task of sentiment analysis on the IMDB movie reviews dataset in Python. Laplace Smoothing was used to handle new and missing words. The Perceptron algorithm was implemented for the same task and feature selection was done using the Bag of Words model. The accuracy was measured by the ten-fold cross-validation technique. The Perceptron algorithm gave a better accuracy as expected
Question Answering Using Deep Learning
February 1, 2017 – April 1, 2017
Implementing various Neural Network models such as RNNs, LSTMs, End-to-End-Memory Networks for the complex task of Question-Answering. This project is implemented using the bAbI dataset in Keras Deep Learning package. After performing the pre-processing of the data, each word was converted to a vector and fed into the above models. RNN performance was taken as the baseline. LSTMs and GRUs performed well. However, End-to-End Memory networks outperformed all other models, especially on complex tasks.
Sentiment Analysis to Classify Movie Reviews using LSTM based Recurrent Neural Network
January 1, 2017 – Present
The IMDB dataset was used to classify movie reviews into positive and negative classes. After pre-processing the data, it was fed into an embedding layer to convert words to word embeddings. These embeddings were fed into a series of LSTM cells and finally, the output was taken via a sigmoid activation function for binary classification. The accuracy obtained was almost 80%.
Cultural Fit Analysis
The candidate's extensive experience at Adobe, progressing through various Machine Learning Engineer roles, indicates a strong fit for a fast-paced, innovation-driven environment. The diverse range of personal projects, from traditional ML to advanced deep learning, shows intellectual curiosity and a continuous learning mindset. However, the target role is 'Data Analyst', which is a significant shift from their current and past 'Machine Learning Engineer' roles. While the candidate possesses strong analytical skills, the core focus of their experience is on ML model development and deployment rather than traditional data analysis, reporting, or business intelligence. This misalignment with the target role impacts cultural fit for a pure data analyst position.
Soft Skills & Operational Fit
The candidate's project descriptions and career progression at Adobe suggest strong problem-solving abilities, a proactive approach to complex technical challenges, and leadership potential. The focus on end-to-end ML pipeline ownership indicates operational readiness. However, without specific psychometric test results, a detailed assessment of stress handling, work attitude, and team collaboration is not possible.