May 2023 – Apr 2025 · 1 yr 11 mos · San Jose, California, United States · Hybrid

• Backend software engineer at Watsonx Orchestrate, developing AI agents for business productivity to handle procurement workflows in Coupa.
• Engineered the AI agent through instruction prompts using Meta's Llama 3.1 405b model.
• Owned the development of complete agentic workflow of purchasing - from creating a purchase request to adding items with supplier and billing up to submitting a purchase order for approval.
• Led the team with tool creation for the AI agents in Python and made the overall development process more efficient by reducing code duplication.
• Developed 15 tools that were used by purchasing AI agent to complete the overall CRUD and approval workflows of purchase requests.
• Researched, explored and validated APIs for Coupa using Postman collection
• Backend product engineer at Watsonx Orders, developing AI to automate order taking at restaurant drive-thru.
• Owned backend of core audio and chat manager microservices for voice-based AI application in Python.
• Designed multi-client system through minimal code impact and rolling updates, acquiring client contract for $2M.
• Led refactoring of legacy chat manager service for improved chat session handling, reducing escalations by 10%.
• Architected new WebSocket based audio and event streaming system, enabling remote QA for >90% cases.
• Optimized headset escalation control flow to prevent agent crosstalk and back-to-back escalations in 5% sessions.

Dec 2021 – Mar 2023 · 1 yr 3 mos · Mountain View, California, United States · Hybrid

• Scaled model training pipelines in C++, Python and R to save training resource cost by >50% through deterministic downsampling of training data for a $1B product.
• Implemented, analyzed and verified prediction and bidding modeling pipelines to ensure no spend in cold start issues for Display Ads campaigns using Python and SQL. Launch saved $1M in wasted revenue.
• Created traffic specific budget experimentation infrastructure enabling 3x larger experiments for our team.
• Optimized daily budget spend by reviewing relevant Ad rejections to generate inventory estimation curves and hourly spend fractions. Programmed parallel data processing pipelines in C++ and code review and debugging in Java.

Apr 2020 – Dec 2021 · 1 yr 8 mos · San Francisco, California, United States

• Leading a team of 4 on a project aimed at automating Electronics part creation by internal content team using Python libraries web service deployment on AWS using Terraform and Github Actions, reducing average part creation times from 10 minutes to 1 minute
• Working alongside architects to implement native command framework in C++ for Fusion Electronics using core command inputs and UX. Aiming to reduce latency and improve rendering performance by >30%.
• Architected and built user message and command events pipeline in C++ through robust API to integrate EAGLE into Fusion 360.
• Added composite commands framework in Fusion Electronics to perform multiple commands through one command.
• Worked on navigation (Design Manager) and PCB routing (Quick Route Guided) features to improve user accessibility, reducing PCB design times by 10%. Developed features on Windows and Unix based systems over Git. Front-end development in Qt, HTML, CSS.
• Implemented object-oriented design for components grouping feature (Persistent Groups). Supported hierarchy & overlap of groups, their selection & operation with 12+ existing commands. Groups were made first class members of the data model.

Jun 2018 – Apr 2020 · 1 yr 10 mos · San Francisco, California, United States

Jan 2018 – May 2018 · 4 mos · San Jose, California, United States · On-site

• Designed and developed generic modules in Python to clean and preprocess financial data from 5+ file formats (PDF/JPG/Excel/CSV/TSV etc.)
• Implemented analytical methods to recognize and correct possible errors in numerical and textual data with more than 90% accuracy
• Developed robust mathematical and statistical models to identify human errors while drafting financial spreadsheets and suggest possible corrections
• Tested developed modules on 200+ files to give feedback to ontology improving mechanism

May 2017 – Aug 2017 · 3 mos · San Francisco, California, United States

• Worked with EAGLE Core Systems Team to conceptualize the feature of Persistent Groups in Eagle.
• Developed a proof of concept which enabled creation of named persistent group of components in circuit schematic.
• Refactored more than 12 existing commands in EAGLE to work on persistent group for easy selection, processing and deletion of saved group of objects.
• Worked on large and complex object-oriented code base, software development lifecycle, Agile Methodologies, and tools like JIRA and Github.
• Ensured development was done within robust frameworks maintaining 25 years old stability of EAGLE code with programming in various languages like C, C++, HTML, CSS, JavaScript, XML and Python.

Jan 2017 – Apr 2017 · 3 mos · Los Angeles, California, United States

• Working in the department of Information Technology Services (Learning Environments) at USC
• Duties include:
• Provide customer support to access over 200 Multimedia Enhanced Classrooms across USC
• Troubleshoot technical problems over telephone and in-person ensuring less than 5-minute class time impact
• Report problems to fellow STAs and Specialists while monitoring classroom activities during operational hours