Pratik

AI/ML Engineer Intern

IndiaNIC • Fulltime

Feb 2026 - Present • 3 mos

AI/ML Engineer Intern at National e-Governance Division (NeGD), Ministry of Electronics & IT, Government of India. Project 1 — AI-Powered Expert Selection System: Built an end-to-end RAG pipeline processing 17,000+ biodata records for a central government recruitment body. Engineered multi-stage pipeline: rule-based SQL filtering → BERT embedding → FAISS semantic retrieval → Random Forest ranking → linear programming for diversity constraints. Reduced panel generation time from several hours to under 5 minutes and cut manual shortlisting effort by 80%. Project 2 — AI Office Management System (Ministry of Youth Affairs & Sports): Built a full LLM-powered office digitization platform for a 40+ person senior government office. Designed LangChain-based agentic workflows for natural language task parsing via WhatsApp Bot. Integrated LangGraph DAG orchestration for multi-step document approval flows. Built LLM inferencing pipeline using Claude API covering 12 official document types. Implemented pgVector for semantic document retrieval, deployed on AWS (EC2, S3, RDS), containerized via Docker with Supabase Edge Functions and pgcron for automated reminders. Boosted task capture rate from 60% to 95% and reduced manual coordination effort by 70%. Built secure APIs with JWT auth and OWASP API security standards serving the Joint Secretary's office.

Software Developer

Government of Uttar Pradesh • Fulltime

Mar 2025 - Jan 2026 • 10 mos

Software Developer at Information and Public Relations Office, Banaras Hindu University, Varanasi, India. (March 2025 – January 2026) Responsibilities and Achievements: Web Platform Development: Led the design and development of institutional web platforms for one of India's largest central universities. Built high-performance websites using Next.js with server-side rendering (SSR), Strapi as a headless CMS, and PostgreSQL as the primary database. Integrated Redis caching layer to optimize repeated queries and static content delivery — reduced average page load time by approximately 40%, significantly improving user experience for thousands of daily visitors across the university's digital infrastructure. NLP & Sentiment Analysis Pipeline: Engineered a complete NLP-based sentiment analysis pipeline from scratch using Python, Pandas, and spaCy to monitor and track public opinion trends around university events, announcements, and initiatives. The pipeline automated data collection, preprocessing, entity recognition, and sentiment classification across multiple text sources. Data Reporting & Dashboards: Designed and delivered interactive Power BI dashboards replacing entirely manual Excel-based reporting processes used by the communications team. This automation saved approximately 70% of reporting effort every week and provided real-time visibility into key metrics for university leadership.