About
About me
I’m a software engineer with 10 years of experience building and scaling web applications, now focused on production AI systems, enterprise integrations, and cloud-native delivery. I work hands-on with Python, AWS, and serverless architectures, and take a practical approach to RAG, LLM workflows, governance, and operational reliability.
I enjoy partnering across teams to turn complex business needs into systems that are usable and maintainable.
Experience
Senior Software Engineer — Bank of America
Chicago, IL · Jul 2025 – Present
Built and supported production software in a fast-moving consulting environment, focused on cloud delivery, workflow automation, and scalable application design that can support AI-enabled capabilities.
- Built Python services on AWS Lambda with API Gateway to replace manual intake steps, cutting turnaround time by 38% and supporting 12,000+ monthly requests.
- Designed integration patterns for serverless architectures and external systems, cutting handoff delays by 41% and moving releases from biweekly to weekly.
- Implemented RAG-ready data flows with vector databases and document preprocessing, improving answer relevance by 29% in internal validation.
- Added LLM evaluation gates, automated tests, and monitoring hooks, lowering production regressions by 33% across 6 major deployments.
- Introduced governance controls, metadata tagging, and access rules, reducing policy exceptions by 45% and improving audit readiness.
- Partnered with product and architecture leads to define AI agent execution paths and integration points, moving 3 pilot workflows into production planning.
Software Engineer — Ad Hoc
Remote · Aug 2021 – Jun 2025
Delivered enterprise software supporting operational workflows and large-scale internal systems, working closely with business and platform teams on reliability, integration quality, and data-driven automation.
- Rebuilt core services in Java and AWS, improving response times by 46% and supporting 20,000+ daily transactions.
- Developed API and service interfaces for workflow automation, reducing manual reconciliation effort by 52% across 4 business groups.
- Added SQL-backed logging, observability dashboards, and event tracking, cutting incident resolution time from hours to under 30 minutes.
- Built transformation logic for enterprise data workflows, improving downstream accuracy by 24% and reducing rework.
Junior Software Engineer — UPS
Parsippany, NJ · Dec 2016 – Aug 2021
Supported application development and maintenance for enterprise systems, contributing to backend services, data handling, and release support in a high-volume environment.
- Built Java and SQL utilities to automate common support tasks, reducing ticket volume by 31% and saving 15+ hours per week.
- Implemented validation and transformation logic in backend services, improving record accuracy by 27% across 3 operational workflows.
- Helped stabilize CI/CD checks and environment setup, cutting failed deployments by 40% over 10 release cycles.
- Added structured logs and observability signals, shortening debugging cycles by 35% for recurring issues.
Skills
- Languages: Python, C#, Java, JavaScript, TypeScript, SQL
- AI & LLM: AWS Bedrock, Amazon SageMaker, LangChain, RAG pipelines, vector databases, agentic workflows, LLM evaluation, prompt engineering
- .NET: .NET 8, ASP.NET Core, MVC, Web API, Razor Pages, Blazor, Entity Framework Core, LINQ, SignalR, xUnit, NUnit, Moq
- Cloud & web: AWS, Lambda, API Gateway, serverless architectures, FastAPI, Node.js, React, REST APIs, pytest
- Systems: Enterprise search, observability, metadata management, lineage tracking, access and governance controls, document intelligence, production monitoring
Projects
- Control (2026) — A local-first mission control platform built with Next.js and SQLite that automates multi-agent workflows through a tag-based dynamic orchestration pipeline, with a model-agnostic assignment system that routes tasks across LLMs by complexity and resource needs.
- Reveal Once (2025) — An end-to-end encrypted secret sharing app built with React and Node.js, featuring cron-based link expiry and burn-after-read one-time sharing.
- Redditlite (2025) — A minimal, high-performance Reddit client built with Next.js, using predictive prefetching, API response caching, and a performance-first UI.
Education
- M.S. in Computer Science, New York Institute of Technology (2018)
- B.S. in Computer Engineering, Ahmedabad Institute of Technology (2015)
If you want to get in touch, email me at harshil_1993@outlook.com.