About Me
Hey, I'm Bhumika. I actively seek out environments where I'm pushed to grow, I'd rather be learning something new every day than staying comfortable doing what I already know. I enjoy figuring things out from scratch, asking the right questions when things aren't clear, and taking ownership of problems end to end.
I started with web development, MEAN and MERN stack, then got curious about machine learning and AI, which led me to pursue my MS in Computer Science at Syracuse University. Now I focus on LLMs and agentic AI , building RAG-based systems, chatbots, Slack bots, and working with MCP. I currently work as a software engineer at Stellar, an AI & data consulting firm, building AI-driven applications for clients across HR and service platforms, automating legacy workflows to make them faster and more efficient.
I like bringing creativity into what I build, but I care just as much about making it efficient, practical, and something that actually works in the real world.
Outside of code, I'm training for my first half marathon in May 2026 🏃♀️. Same approach: show up, stay consistent, keep going even when it's hard.
Experience
Software Engineer
● NOWStellar
Full-time · Hybrid
- ▸Built an AI-powered analytics feature that automatically reads through Slack bot conversations and surfaces the top trending HR topics to admins, replacing hours of manual log review with an instant, filterable dashboard.
- ▸Built an MCP server that lets external AI clients (like Claude) securely connect to and query the platform's HR data via 10 tools across 5 domains, making the AI assistant more context-aware and useful for HR teams.
- ▸Built a document management system that lets companies upload large policy and compliance documents, processes them using LangChain and RAG (chunking, embedding, vector search), and exposes them as AI tool calls so the Slack bot can accurately answer employee questions grounded in real company knowledge.
- ▸Owned the full-stack jobs system end-to-end on a home services platform, designing the DB schema, REST API, admin dashboard, contractor and employee assignment workflows, AI tool integrations, and PDF/Excel ROI reporting, across multiple user roles and surfaces.
- ▸Built a multi-tenant white-label partner portal for brokerages and realtors, including RBAC-gated access, branded public pages, group invite flows with email automation, a client relationship dashboard, and Stripe-linked trial signups tied to partner identity.
- ▸Engineered a homeowner property management hub with relational data across appliances, documents, emergency contacts, and a home service team, with a shareable public view, replacing entirely manual, fragmented workflows across spreadsheets and emails.
- ▸Contributed fixes and improvements across login flows, document uploads, mobile experience, and messaging reliability on both platforms.
Software Developer
Kollabio, Inc.
Part-time · Remote
- ▸Designed and integrated AI-powered backlog generators using Mistral and GPT-3.5, enhancing efficiency of backlog creation within the KollabKubes platform.
- ▸Managed and upgraded multiple LoopBack APIs; integrated Vue.js with Nest.js for seamless data flow and Dockerized deployment with MySQL databases.
- ▸Collaborated closely with cross-functional teams to troubleshoot complex issues and optimize project workflows.
GIS Software Engineer
Syracuse University · SU OSPO
Part-time · Hybrid
- ▸Developed an open-source Python-based QGIS plug-in for spatial data visualization, enabling users to search and select map icons from a library of 1,000+ assets.
- ▸Implemented a structured metadata framework to improve spatial data analysis and user accessibility.
- ▸Designed and refined the plug-in's UI/UX through iterative feedback from cartography and GIS experts.
Graduate Research Assistant
Syracuse University · iSchool
Part-time · On-site
- ▸Facilitated research on digital literacy and equity through community outreach and workshops aimed at bridging the racial wealth gap.
- ▸Collaborated with open-source communities including OpenCommons, NIST, and Smart Cities to integrate OSS into research initiatives.
- ▸Represented Syracuse University at UN OSPO'24, advocating for open-source solutions and digital literacy.
Summer Intern
Citi
Internship · Remote
- ▸Built an internal web application with Angular and Spring Boot, enabling 100+ employees to access domain-specific APIs via a unified dropdown interface.
- ▸Integrated Redis caching, reducing API response times by 30% and improving performance for non-technical users.
- ▸Designed and deployed RESTful APIs to dynamically render domain-specific data, cutting server load by 25%.
Currently Building
LinkedIn Post Generator
I post regularly on LinkedIn about what I'm learning and building, but writing a good post from scratch every time is slow. So I'm building a tool to do it better.
It connects to my Notion and GitHub via MCP to automatically pull in context about what I've been working on. I can also attach research papers or articles as a source of truth. From there, I prompt it to draft a post, refine the tone, and save the prompts I liked for future use, so I'm not starting from zero every time.
How it works
Pull context
Connects to Notion and GitHub via MCP to fetch recent work, commits, and notes automatically.
Attach sources
Link research papers or articles as grounding material so the AI writes from real context, not hallucination.
Generate and refine
Prompt the LLM to draft a LinkedIn post, iterate on tone and structure until it sounds like you.
Save prompts
Store the prompts that worked so you can reuse and improve them over time.
What I'm thinking about next
I want to keep building tools that solve real friction in my own workflow, especially at the intersection of agentic AI and everyday productivity. The pattern I keep coming back to: give an AI the right context, the right tools, and a clear goal, and it becomes genuinely useful, not just impressive. I'm also exploring how MCP can be used to build AI workflows that are composable, reusable, and actually integrate with the tools people already use.
Got a fun idea? Let's build it together.
I'm always up for a conversation, a side project, a collab, or just talking through an idea.
Projects
Multimodal Meme Classification
I wanted to understand how AI can make sense of images and text together — so I built a system to classify memes for harmful content. Turns out combining vision models and language embeddings is messier than it sounds, and that's what made it interesting.
Sage – First Aid Bot
Curious about what it actually takes to make an LLM useful — not just smart, but practically helpful. Built Sage using Mistral 7B to give real first-aid guidance, and spent as much time on the UX as the model.
AI-Powered Learning Path Generator
Explored whether AI could actually personalize learning — not just recommend courses, but build a structured path around someone's background and goals. The result is a tool that outputs a real, downloadable plan.
Job Information Tracker
Built this out of frustration with my own job search — I kept losing track of applications. A lightweight Chrome extension to capture and organize job info as you browse. Showcased at the Google Chrome Built-in AI Challenge.
Plant Disease Detection
Wanted to explore image classification with a real-world impact angle. Built a system that lets farmers photograph a plant leaf and get an instant disease diagnosis — practical ML, not just a Jupyter notebook.
Cold Message Generator
A tool I wished existed during my own job hunt — paste a job URL, get a tailored cold message. Chained web scraping with an LLM to do in seconds what used to take 20 minutes of manual writing.
What if attention data from your brain could tell you which platform you actually learn best on? That question drove this research project — using EEG signals to recommend e-learning platforms based on how engaged you genuinely are.
My first real full-stack project — I wanted to understand how all the pieces of a web app fit together. Built a buy/sell marketplace for second-hand books and learned more from the bugs than the features.
Built a blogging platform for my college because we didn't have one. Gated access behind college email IDs so the community stayed authentic. First time I really thought about who I was building for.
Learning React by building something simple but interactive — focused on getting state management and component communication right before moving to bigger things.
Built to get comfortable with React refs and portals — the parts of React that don't show up in tutorials until you actually need them.
A small tool I actually use — projects compound growth so you can see what consistent investing looks like over time. Simple to build, satisfying to use.
13 of 13 projects
More on GitHub →Skills
Languages
Frontend
Backend & APIs
AI / ML
Databases
Cloud & DevOps
GitHub Activity
Education
Syracuse University
Master of Science, Computer Science
2023 – 2025 · Syracuse, NY
- ▸Graduate Research Assistant — iSchool
- ▸GIS Software Engineer — SU OSPO
- ▸Open-source community contributor · UN OSPO'24
SSN College of Engineering
B.E., Computer Science & Engineering
2019 – 2023 · Chennai, India
- ▸EEG-based e-learning recommendation research
- ▸SSN Blog platform (Flask)
- ▸Competitive programming — LeetCode, Coding Ninjas
Get In Touch
I'm always open to interesting conversations — a project idea, a potential role, or just a chat about building things. My inbox is open.
Say Hello →