Open to opportunities

archita sharma

@architasharma1

Message

Self-taught machine learning engineer with 3+ years building NLP, computer vision, and predictive analytics solutions.

Zimbabwe

Message

What I'm looking for

I’m looking for fully remote work where I can build and deploy production ML—especially NLP, computer vision, and time series—using strong MLOps practices (monitoring, retraining, drift detection). I want teams that value continuous learning and measurable impact.

I’m a self-taught machine learning engineer with 3+ years of hands-on experience building and deploying ML solutions at Fortune 500 companies. I transitioned from literature studies to AI/ML through rigorous self-education, intensive practice, and continuous learning.

At Sakon (TEM AI Platform), I develop and deploy 6+ NLP models for anomaly detection, fraud classification, and automatic categorization—improving accuracy to 99.2% and reducing manual revisions by 60%. I also built an invoice categorization pipeline with BERT/transformer models that processes 10,000+ invoices daily with a 35% error reduction, and I engineered 200+ dynamic features to raise F1 from 0.78 to 0.92.

I further optimize model performance and delivery by reducing inference latency by 40% using XGBoost and LightGBM, deployed as an AWS Lambda microservice for real-time predictions. I’ve also implemented an end-to-end MLOps pipeline with monitoring (Evidently AI), automated retraining, and drift detection, reducing incidents by 50%.

Previously at Numerator, I led 5+ end-to-end projects in time series classification, clustering, and forecasting with 100% on-time delivery and high-precision benchmarks, including an ARIMA + Prophet demand forecasting model achieving MAPE of 8.5% and saving ₹2.3M in annual inventory costs. Earlier at Amazon, I delivered a computer vision product image classifier (CNN ResNet50) with 94% accuracy and built a scalable Python + Apache Spark pipeline for preprocessing and feature extraction at 100GB+ scale.

Experience

Work history, roles, and key accomplishments

Current

Associate ML, Platform Intelligence

Current

Sakon

Sep 2024 - Present (1 year 7 months)

Developed and deployed 6+ NLP models for telecom-bill anomaly detection, fraud classification, and auto-categorization, reaching 99.2% accuracy and cutting manual revisions by 60%. Built an invoice categorization pipeline (BERT/transformers) processing 10,000+ invoices daily, reducing error by 35%, and improved model F1 from 0.78 to 0.92 while reducing inference latency by 40% via AWS Lambda deplo

NLP Anomaly Detection Transformers XGBoost LightGBM AWS Lambda A B Testing

Current

Associate ML Engineer

Current

Sakon

Sep 2024 - Present (1 year 7 months)

Developed and deployed 6+ NLP models for anomaly detection, fraud classification, and automatic categorization, reaching 99.2% accuracy and reducing manual revisions by 60%. Built an invoice categorization pipeline (BERT/transformers) processing 10,000+ invoices/day with 35% lower error, engineered 200+ features to improve F1 (0.78→0.92), and implemented MLOps with Evidently AI to cut incidents by

NLP BERT Transformers Anomaly Detection XGBoost AWS Lambda

Data Scientist, Predictive Analytics

Numerator

Jun 2023 - Sep 2024 (1 year 3 months)

Led 5+ end-to-end predictive analytics projects (time series classification, clustering, and forecasting) with 100% on-time delivery and strong benchmark performance. Built an ARIMA + Prophet demand forecasting model with 8.5% MAPE and delivered ₹2.3M annual inventory savings, plus feature engineering across 500K+ retail sales tickets.

Time Series ARIMA Prophet Clustering Data Cleaning Python Statistical Validation

Data Scientist

Numerator

Jun 2023 - Sep 2024 (1 year 3 months)

Led 5+ end-to-end predictive analytics projects for time series classification, clustering, and forecasting, delivering 100% on-time with high-precision benchmarks. Built a demand forecasting model (ARIMA + Prophet) with 8.5% MAPE, saving ₹2.3M annually in inventory costs, and engineered 150+ features from 500K+ retail sales tickets to capture seasonality, promotions, and competitive dynamics.

Time Series ARIMA Prophet Clustering Data Cleaning SQL Python

ML Associate, Marketplace ML

Amazon

Apr 2022 - May 2023 (1 year 1 month)

Built a CNN ResNet50 computer-vision model for product image classification on the French marketplace, achieving 94% accuracy on 50K+ images. Developed a scalable Python + Apache Spark pipeline to preprocess and augment images, processing 100GB+ with 99.5% data integrity.

Computer Vision CNN ResNet50 Apache Spark Data Pipelines data preprocessing Python Feature Extraction

ML Associate (Computer Vision)

Amazon

Apr 2022 - May 2023 (1 year 1 month)

Built a ResNet50 CNN computer vision model for product image classification on the French marketplace, achieving 94% accuracy on 50K+ images. Developed a scalable Python + Apache Spark preprocessing and augmentation pipeline processing 100GB+ data with 99.5% integrity.

Computer Vision ResNet50 CNN Apache Spark Feature Extraction Python OpenCV Data Pipelines

Education

Degrees, certifications, and relevant coursework

Symbiosis Institute of Management

Master of Business Administration (MBA), Business Administration

Activities and societies: Self-taught ML through 15+ courses (Deep Learning Specialization, MLOps, NLP Transformers, Computer Vision, Advanced SQL); Kaggle practice; open-source contributions.

Part-time MBA studies while completing extensive self-taught ML coursework, including deep learning, MLOps, NLP transformers, and computer vision. Actively practices on Kaggle and contributes to open-source.