Machine Learning Engineer at RegCore AI

Engineered end-to-end retrieval-augmented generation system for financial institutions featuring advanced query processing, custom chunking algorithms, and hybrid search (BM25 + vector similarity), resulting in 40% improved response accuracy over traditional chatbot solutions. Built production-scale data ingestion pipeline with automated web scraping, real-time preprocessing, and PostgreSQL + pgVector architecture using HNSW indexing, processing 10K+ financial documents with 99% uptime and successfully delivering client demonstration to Forrester Research.

Machine Learning Engineer at XRI Global AI

Implemented cross-lingual vocabulary transfer for adapting Mistral-7B models to new languages, achieving 87% accuracy in Hindi-English translation. Implemented token alignment using SMT tools and fine-tuned LLMs on low-resource languages to optimize token mapping and performance across Latin and non-Latin scripts. Developed a Full-Stack Interactive Language Technology Dashboard to visualize African language data, integrating interactive charts (BLEU, chrF++), geospatial mapping, and RESTful APIs for data aggregation, using PySpark for processing. Improved data accessibility by 40% and ensured database reliability with automated backups.

Machine Learning Engineer at ProjectX.Cloud

RAG System Development: Built a production-ready retrieval-augmented generation (RAG) chatbot using Python, FastAPI, LangChain, and Chroma DB, trained on 1,000+ documents. Implemented RESTful API endpoints to handle user queries with Redis-based caching and Celery for background task processing. Impact: Enhanced chat response time by 1.5x through RAG optimization and enabled the PR agent to detect 80% of code quality issues, supporting a prototype platform handling 500 daily user queries before discontinuation due to strategic challenges. Performance improvements included caching strategies and asynchronous processing to improve API response times by 45%, ensuring scalability for multiple concurrent language processing requests.

Machine Learning Engineer at RegCore AI

RAG Application for Banks: Engineered end-to-end retrieval-augmented generation system for financial institutions featuring advanced query processing, custom chunking algorithms, and hybrid search (BM25 + vector similarity), achieving a 40% improvement in response accuracy over traditional chatbot solutions. Pilot with banks and other firms: Built production-scale data ingestion pipeline with automated web scraping, real-time preprocessing, and PostgreSQL + pgVector architecture using HNSW indexing, processing 10K+ financial documents with 99% uptime and delivered client demonstration to Forrester Research.

Machine Learning Engineer at XRI Global AI

LLM Trans-tokenization: Developed a cross-lingual vocabulary transfer strategy for adapting Mistral-7B models to new languages, achieving 87% accuracy in Hindi-English translation. Implemented token alignment using SMT tools and fine-tuned LLMs on low-resource languages to optimize token mapping and performance across Latin and non-Latin scripts, with improved BLEU, Perplexity and CHRF scores. Dashboard and Data Integration: Built a Full-Stack Interactive Language Technology Dashboard to visualize African language data, integrating interactive charts (BLEU, chrF++), geospatial mapping (Leaflet.js), and RESTful APIs for data aggregation; processed with PySpark, improving data accessibility by 40% and ensuring automated backups.

Machine Learning Engineer at ProjectX.Cloud

RAG System Development: Built a production-ready retrieval-augmented generation (RAG) chatbot using Python, FastAPI, LangChain, and Chroma DB, trained on 1,000+ documents. Implemented RESTful API endpoints to handle user queries with Redis-based caching and Celery for background task processing. Impact: Enhanced chat response time by 1.5x through RAG optimization and enabled the PR agent to detect 80% of code quality issues, supporting a prototype platform handling 500 daily user queries. Performance Optimization: Implemented caching strategies and asynchronous processing to improve API response times by 45%, ensuring scalability for multiple concurrent language processing requests.

Machine Learning Engineer at XRI Global AI

Machine Learning Engineer responsible for cross-lingual LLM adaptation and dashboard data integration. Developed a cross-lingual vocabulary transfer strategy for adapting Mistral-7B models to new languages, achieving 87% accuracy in Hindi-English translation. Implemented token alignment using SMT tools and fine-tuned LLMs on low-resource languages to optimize token mapping and performance across Latin and non-Latin scripts. Built a Full-Stack Interactive Language Technology Dashboard with interactive charts (BLEU, chrF++), geospatial mapping (Leaflet.js), and RESTful APIs for data aggregation; used PySpark for processing. Improved data accessibility by 40% and ensured database reliability with automated backups.

Machine Learning Engineer at ProjectX.Cloud

RAG System Development: Built a production-ready retrieval-augmented generation (RAG) chatbot using Python, FastAPI, LangChain, and Chroma DB, trained on 1,000+ documents. Implemented RESTful API endpoints to handle user queries with Redis-based caching and Celery for background task processing. Impact: Enhanced chat response time by 1.5x through RAG optimization and enabled the PR agent to detect 80% of code quality issues, supporting a prototype platform handling 500 daily user queries before discontinuation due to strategic challenges. Performance Optimization: Implemented caching strategies and asynchronous processing to improve API response times by 45%, ensuring scalability for multiple concurrent language processing requests.

Machine Learning Engineer at Visionway

AI-powered interview software with a client-side cheating detection dashboard; trained ML models in Python using facial landmark analysis for eye and head tracking on a dataset of 500 images; integrated OpenAI API for question generation and answer scoring; results displayed via a JSON-based FastAPI dashboard using Redis and Celery for caching. Backend Architecture: Designed and implemented a scalable backend system using Python, FastAPI, Redis and Celery that processed real-time video streams and AI analysis results. Created RESTful API endpoints that handled 100+ daily requests while maintaining sub-second response times.

Machine Learning Engineer at RegCore AI

RAG Application for Banks: Engineered end-to-end retrieval-augmented generation system for financial institutions featuring advanced query processing, custom chunking algorithms, and hybrid search (BM25 + vector similarity), resulting in 40% improved response accuracy over traditional chatbot solutions. Pilot with banks and other firms: Built production-scale data ingestion pipeline with automated web scraping, real-time preprocessing, and PostgreSQL + pgVector architecture using HNSW indexing, processing 10K+ financial documents with 99% uptime and successfully delivering client demonstration to Forrester Research.

Machine Learning Engineer at XRI Global AI

LLM Trans-tokenization: Developed a cross-lingual vocabulary transfer strategy for adapting Mistral-7B models to new languages, achieving 87% accuracy in Hindi-English translation. Implemented token alignment using SMT tools and fine-tuned LLMs on low-resource languages to optimize token mapping and performance across Latin and non-Latin scripts. Dashboard and Data Integration: Developed a Full-Stack Interactive Language Technology Dashboard to visualize African language data. Integrated interactive charts (BLEU, chrF++), geospatial mapping (Leaflet.js), and RESTful APIs for data aggregation, using PySpark for processing. Improved data accessibility by 40% and ensured database reliability with automated backups.

Machine Learning Engineer at ProjectX.Cloud

RAG System Development: Built a production-ready retrieval-augmented generation (RAG) chatbot using Python, FastAPI, LangChain, and Chroma DB, trained on 1,000+ documents. Implemented RESTful API endpoints to handle user queries with Redis-based caching and Celery for background task processing. Impact: Enhanced chat response time by 1.5x through RAG optimization and enabled the PR agent to detect 80% of code quality issues, supporting a prototype platform handling 500 daily user queries. Performance Optimization: Implemented caching strategies and asynchronous processing to improve API response times by 45%, ensuring scalability for multiple concurrent language processing requests.

Machine Learning Engineer at RegCore AI

Designed and deployed an end-to-end retrieval-augmented generation (RAG) system for banks, featuring advanced query processing, custom chunking, and hybrid BM25 + vector search. Achieved a 40% improvement in response accuracy vs traditional chatbot solutions. Built a production-grade data ingestion pipeline with automated web scraping, real-time preprocessing, and a PostgreSQL + pgVector architecture using HNSW indexing. Managed 10K+ financial documents with 99% uptime and delivered client demos to Forrester Research.

Machine Learning Engineer at XRI Global AI

Developed a cross-lingual vocabulary transfer for adapting Mistral-7B models to new languages, achieving 87% accuracy in Hindi-English translation. Implemented token alignment using SMT tools and fine-tuned LLMs on low-resource languages to optimize token mapping and performance across Latin and non-Latin scripts. Built a full-stack language technology dashboard (BLEU, chrF++, Leaflet.js) with RESTful APIs and PySpark processing; improved data accessibility by 40% and ensured automated backups.