Senior AI/ML Engineer at Kesh Tec INC (2023-01 – 2025-08)
Designed, developed, and deployed advanced AI/ML pipelines, Python backend systems, and Next.js applications for healthcare AI, model integration, clinical automation, and enterprise analytics.
- Built Python ETL pipelines to ingest and process structured healthcare data, including claims and labs, as well as unstructured data such as clinical notes and imaging records.
- Applied NLP techniques including tokenization, embeddings, transformers, and clinical entity extraction to support healthcare data intelligence.
- Developed HIPAA/GDPR-compliant preprocessing workflows, including de-identification pipelines for sensitive clinical data.
- Fine-tuned transformer-based models, including BERT and GPT-family models, on EHR and patient communication data to predict appointment no-shows and care drop-offs.
- Created longitudinal patient features by combining time-series health data with demographic and historical records.
- Built Next.js dashboards to visualize patient risk scores, appointment trends, and radiology AI outputs.
- Exposed machine learning risk scores through Python REST APIs on AWS Lambda and integrated them with scheduling and outreach systems.
- Trained CNNs and Vision Transformers for radiology image classification and segmentation tasks.
- Built GPU-accelerated inference pipelines for real-time DICOM image processing and deployed imaging models into PACS-integrated systems using ONNX Runtime.
- Built medical chatbots using RAG pipelines based on clinical guidelines and hospital knowledge bases.
- Deployed voice-enabled virtual assistants for patient outreach and triage.
- Implemented daily quality monitoring using AWS CloudWatch and human-in-the-loop annotation workflows.
Senior Python Engineer at Kesh Tec Inc (2020-02 – 2022-12)
Built high-performance Python, machine learning, and financial data systems for fraud detection, credit risk, document intelligence, trading analytics, and regulatory reporting.
- Built real-time ML pipelines processing billions of transactions per month using Python, Spark Streaming, Kafka, and AWS Kinesis.
- Developed anomaly detection models using deep autoencoders and ensemble methods, then deployed fraud scoring APIs with sub-200ms latency.
- Designed Next.js reporting tools to surface fraud alerts and compliance dashboards for risk officers.
- Engineered model features from customer profiles, credit history, transaction activity, and behavioral data.
- Built explainable ML models using XGBoost and SHAP values to support regulatory transparency requirements.
- Delivered Python REST APIs for credit risk scoring consumed by loan origination platforms.
- Applied NLP to contracts, earnings reports, filings, and financial documents for entity extraction and sentiment analysis.
- Fine-tuned transformer-based models for financial text classification.
- Automated document ingestion pipelines using OCR, Tesseract, and AWS Textract.
- Built time-series forecasting models for price prediction and volatility forecasting.
- Designed Python backtesting and simulation pipelines for trading strategies.
- Integrated financial systems with trading platforms using FIX and custom APIs.
AI Software Engineer at Quip Global (2018-09 – 2020-01)
Developed Python-based ML and NLP pipelines for enterprise SaaS, retail, and e-commerce systems while building React/Redux frontends for personalization dashboards and customer-facing AI features.
- Built retrieval-based pipelines in Python for enterprise knowledge bases.
- Fine-tuned word embeddings and classical NLP models, including LSTMs and BiLSTMs, for domain-specific language tasks.
- Designed multi-turn chatbot workflows for customer support automation.
- Built React/Redux dashboards to visualize ticket trends, model outputs, and bias/fairness audit results.
- Developed ML pipelines to auto-classify tickets, documents, and support queries.
- Integrated APIs with SaaS platforms to trigger automated responses and escalations.
- Built document summarization and Q&A models using classical NLP techniques.
- Conducted bias and fairness audits with automated flagging and review workflows.
- Developed monitoring systems to track prediction drift and model errors.
- Containerized ML services with Docker and deployed batch and API-based inference pipelines for internal and customer-facing applications.
- Automated model retraining workflows using CI/CD pipelines on AWS EC2 and on-premise servers.
Python Backend Engineer at Quip Global (2017-01 – 2018-08)
Built Python backend systems, recommendation engines, NLP pipelines, and real-time personalization services for e-commerce and SaaS applications.
- Built collaborative filtering and deep learning-based recommendation systems serving millions of customers.
- Optimized real-time personalization pipelines for checkout, search, and product discovery experiences.
- Developed NLP pipelines for query understanding and semantic search.
- Deployed Elasticsearch and ML re-rankers to enhance product discovery.
- Built forecasting models for inventory planning using ARIMA, Prophet, and LSTMs.
- Designed dynamic pricing models using demand elasticity and competitor signals.
- Delivered batch and streaming APIs on AWS Lambda and Kubernetes, consumed by React-based product dashboards.
- Automated feature generation pipelines from customer behavior data.
- Integrated customer segmentation outputs into CRM and marketing automation tools.
Software Engineer at Amazon (2015-11 – 2016-12)
Developed data-driven recommendation, NLP, campaign optimization, and reporting systems for advertising, social media analytics, and user engagement platforms.
- Built predictive models for user engagement, ad click-through, and content personalization using Python, Scikit-learn, and Spark MLlib.
- Developed collaborative filtering and matrix factorization pipelines using Surprise and Spark MLlib to deliver personalized recommendations at scale.
- Tuned and validated models using A/B testing and offline evaluation metrics, including precision, recall, and RMSE.
- Extracted insights from social media posts, reviews, and user comments using NLTK, Gensim, Word2Vec, Python, and regex.
- Built sentiment scoring and topic modeling pipelines using LDA, NMF, and Scikit-learn to categorize user opinions and detect emerging topics.
- Automated text preprocessing workflows including tokenization, stopword removal, and lemmatization for large-scale datasets.
- Designed ML models for ad targeting, budget allocation, and campaign optimization.
- Built dashboards and automated reporting pipelines with Tableau, Flask, and REST APIs to track campaign performance and provide actionable insights.
- Conducted cohort analysis, feature engineering, and regression modeling to refine ad placement strategies.
