Loading portfolio…
Tech Guy
I'm a full-stack engineer who also builds AI systems — I've worked across the entire stack from frontend to deployment, but what I'm most drawn to is building AI that actually changes how people work. At my last role, that meant automating investment research for analysts. At Northeastern, that means researching how to make LLMs more reliable.
Hire me as a best team player
Built AI integrated data processing pipelines using Python, LangChain, and OpenAI API to automate code analysis and document extraction across investment research workflows, reducing time of manual processing by around 40% and supporting due diligence on 300+ companies.
Built ML powered evaluation pipelines using TensorFlow and scikit-learn (SVM, Random Forest) to assess code quality and model performance, generating investment reports that helped prevent $2M+ in potential losses.
Designed multi agent Python scripts that parsed product documentation and specification sheets to auto-generate validation rules, transforming a manual QA configuration process into a repeatable automated pipeline.
Deployed document conversion tool using LangChain RAG pipelines to transform unstructured reports (PDFs, whitepapers) into structured datasets in PostgreSQL, improving research accuracy by 35%.
Engineered backend services with Next.js and Redis caching to support high-traffic e-commerce operations, handling 5K concurrent users with 99.5% uptime through load balancing and connection pooling.
Built intelligent data processing workflows integrating scikit-learn classifiers (Naive Bayes, SVM) with SQL data stores and internal tools, automating end-to-end ticket categorization from ingestion to resolution routing.
Deployed serverless data processing functions on AWS Lambda, enabling scalable execution of analysis workloads and reducing infrastructure management complexity.
Developed RESTful APIs using Java Spring Boot with JWT authentication to power an internal Deal Flow Platform, enabling investment teams to track and manage 300+ annual deals across the pipeline.
Built interactive React dashboards with Redux state management providing real-time pipeline visibility across deal stages, reducing reliance on manual status meetings and improving team decision-making speed.
Established CI/CD pipeline using Jenkins with JUnit/Mockito test automation and SonarQube quality gates, reducing deployment cycles from hours to under 20 minutes.
Northeastern University
ITM (SLS) Baroda University
Developed a real-time guardrail system using Python, PyTorch, and Hugging Face Transformers that monitors LLM hidden state representations to detect and prevent sycophantic behaviour, achieving 85%+ flip prediction accuracy across 4 open-source models (Llama, Mistral, DeepSeek, Gemma). Engineered layer-wise probing pipeline using scikit-learn classifiers on Transformer activations to detect confidence degradation patterns 2-3 turns before model output changes, moving from reactive to predictive AI safety.
Built multi-agent system using Python and LangChain that autonomously analyzes slow SQL queries, reasons over schema structure and execution plans, and generates optimized query rewrites.
Full-stack platform for Northeastern students to connect via real-time video calls and messaging, inspired by personal experience with social isolation as an international student. Built with WebRTC video, JWT authentication, student ID verification, and interest-based discovery. Deployed on Netlify, Render, and MongoDB Atlas.
Let's connect
Transform your organization with proven leadership expertise and strategic vision.