Wilson Vidyut Doloy
Software Engineer with over 3 years of experience building AI-driven and full-stack applications that solve real-world problems at scale. Currently at Deloitte USI, I’ve delivered high-impact solutions like a semantic search system, GPT-powered PDF redaction APIs and AI-assisted data extraction workflows.
Proficient in Python (FastAPI), JavaScript/TypeScript (React, Node.js), and SQL/NoSQL databases, I specialize in designing scalable APIs, efficient database schemas, and intuitive, user-centric UIs. I enjoy working end-to-end from concept to deployment, leveraging cloud services like Azure and modern dev practices (CI/CD, containerization, clean architecture).
Passionate about emerging technologies like Generative AI and ML, I’ve built multiple real-world projects integrating AI models, computer vision, and automation. I thrive in fast-paced, mission-driven environments where innovation meets impact.
Experience
Software Engineer
Deloitte USI
Hyderabad, India
- Architected a unified account-change event stream & cursor-based resynchronization framework, reducing downstream integrations from 3 topics to 1, cutting replay times by 60%, & lowering DB CPU utilization by 35%.
- Optimized large-scale data transformation pipelines through workflow parallelization, memory-efficient processing, and elimination of redundant computations, reducing runtime by 5% & peak memory usage by 42% (12 GB to 7 GB).
- Integrated Google ADK agent workflows across three business processes, automating 67 user scenarios and increasing overall pilot process completion to 92%.
- Developed a voice-driven organizational chart generator using Azure Speech AI, d3.js, and GPT-4o with a Python backend, reducing processing time by 25%.
- Enhanced backend logic for a high-traffic API, improving throughput (3 to 5 req/s) & reducing P95 latency by 28%.
- Engineered a high-performance FAISS-based semantic search system for 62M+ records, leveraging GPU acceleration to achieve a 10× latency reduction.
- Improved tax-document extraction efficiency by 30% by designing a hybrid Azure Form Recognizer + RAG + GPT workflow aligned with Excel-based templates.
- Accelerated output generation time by 12× (240 minutes to 20 minutes) using multiprocessing and multi-threading techniques
Web Developer Intern
Autring - SIMMI Foundation
Remote
- Implemented major front-end features for The Autring news platform using React, creating component architecture, routing, and UI improvements that led to a 25% rise in user engagement
- Designed and integrated Node.js APIs for article retrieval, category filtering, and search, handling 300 requests/day and improving response times through pagination and efficient routing
Education
Vellore Institute of Technology
Bachelor of Technology
GPA: 8.73/10.00, Chennai, India
Skills
Projects
Self-Improving Document Extraction Agent
Generative AI, LLM Agents
- Designed a dual-agent, feedback-driven extraction system using GPT-5 and Azure Document Intelligence, enabling automatic prompt improvement and minimizing manual rule updates.
- Built a Streamlit interface with versioned prompt storage, enabling real-time AI model adjustments during a hackathon, securing a top 10 placement among 50+ competing teams and positive feedback from judges.
Achievements
Winner, Deloitte Code-a-thon 2026
Built an AI-powered coach–coachee conversation assistant
Runner-up, Deloitte Tax AI Challenge 2025
Runner-up, Deloitte Code-a-thon 2025
Developed an AI-powered tax diagnostics tool