Rohith Perumandla

Hey there — I'm Rohith. I love building things with AI. What started as curiosity has become an obsession: I turn ideas into production-ready AI systems, end-to-end.

4+ years building ML pipelines, LLM agents, and full-stack AI platforms. Shipped a healthcare AI SaaS solo in under 4 months. Built a macOS/iOS productivity app. Currently building a multi-layered AI automation platform for aesthetics. I do the whole stack — model, backend, frontend, deployment.

I build multi-agent AI systems that are production-grade — with guardrails, security, evaluation pipelines, observability, cost optimization, and scalability baked in from day one. Not demos. Real systems that handle real users.

MS in Data Science from DePaul, where I also did published research on conversational AI for dementia detection. Previously Data Scientist at SetuServ building NLP pipelines at scale.

Always building. Currently looking for my next full-time AI Engineering role.

Currently working on

Experience

Founding Engineer (Vector Referral)

R&D Venture Studio · Part-time

Sep 2025 - Apr 2026 · 8 mos

Chicago, Illinois, United States · Remote

I developed a digital-first referral management platform designed to modernize the dental care ecosystem and am currently testing it with industry advisors. The platform bridges the gap between general practitioners and specialists by using AI to replace fragmented workflows with a seamless, end-to-end digital experience.

Key Responsibilities & Impact

• Greenfield Product Development: Leading full-lifecycle ownership of a platform that reimagines patient discovery, booking, appointment status tracking, and visit management.
• Intuitive UI/UX Design: Designing high-fidelity, interactive components that reduce administrative friction and improve diagnostic clarity.
• Cross-Functional Collaboration: Working closely with dentists, patients, and insurers to align clinical requirements with a modern, user-centric interface.
• Strategic Growth: Using continuous feedback loops and data insights to expand dental care access, increase pricing transparency, and improve long-term oral health outcomes.
• Rapid Iteration: Driving fast cycles of discovery and validation to ensure the roadmap stays ahead of the evolving needs of modern dental practices.

... more

DePaul University Jarvis College of Computing and Digital Media

Part-time · 2 yrs 4 mos

Chicago, Illinois, United States

Machine Learning Engineer

Sep 2023 - Dec 2025 · 2 yrs 4 mos

On-site

As an AI/ML Research Engineer at DePaul University, my focus revolves around building innovative conversational speech systems designed specifically for dementia care. Through the integration of advanced Large Language Models like Llama3 and Phi3, these systems aim to engage patients in meaningful dialogues while monitoring their cognitive health. I led the cross-functional team in the technical development of the project.

One of the proudest moments was developing a module that tracks cognitive status by analyzing linguistic biomarkers; this work has been crucial in providing healthcare professionals with real-time insights into patient conditions. See Publications

Additionally, I created a Novel Topic Detection feature that personalizes conversations based on individual patient interests. The EMA Conversation Module I implemented allows for immediate cognitive assessments during interactions. My passion lies in pushing the boundaries of technology to create empathetic AI solutions that genuinely enhance lives.

Large Language Models (LLMs), Natural Language Processing (NLP), Machine Learning, Deep Learning, Conversational AI

... more

Teaching Assistant

Sep 2024 - Jun 2025 · 10 mos

Remote

Vice President

Jan 2025 - Jun 2025 · 6 mos

On-site

Data Scientist

SetuServ · Full-time

Jan 2022 - Jul 2023 · 1 yr 7 mos

Hyderabad, Telangana, India · On-site

• End-to-End NLP Pipelines: Developed an extensive Data Science pipeline focused on extracting valuable insights from product reviews across numerous e-commerce websites.
• Advanced Sentiment Analysis: Implemented advanced Natural Language Processing models (including pre-trained BERT) that significantly improved our ability to gauge customer sentiment.
• Scraping Automation: Deployed over 100 custom-built web scrapers using Python frameworks like Scrapy and BeautifulSoup, streamlining data collection.
• Efficiency Gains: Reduced manual data collection effort by nearly 50% and achieved a remarkable increase in overall data extraction efficiency by 40%.
• Competitive Intelligence: Collaborated with stakeholders to translate business objectives into actionable insights, directly enhancing competitive intelligence for major e-commerce players.

Machine Learning Engineering, Pre-trained BERT NLP Models, Data Pipelines, Web Scraping Automation

... more

Freelance

Upwork · Freelance

Sep 2022 - Jun 2023 · 10 mos

Hyderabad, Telangana, India · Remote

Research Work

We’re pushing the boundaries of dementia care using AI. For the past 2 years, our team has been building a Conversational Speech System for robots and various assistive modules to support both dementia patients and their caregivers. You can read our recent publication on arXiv: Enhancing Dementia Care with AI .

Our research began with the development of a conversational AI Speech system using NLP techniques. I fine‑tuned open‑source models (Llama 3 and Phi-3 7B ) on the cleaned and preprocessed DementiaBank and Indiana dataset to power natural, context‑aware responses.

To make interactions more engaging and human‑like, we implemented several specialized modules:

• Topic‑Shift Awareness & Triggering: Decides when to introduce a new topic or remain on the current one
• Turn‑Taking Controller: Manages speaker changes for smooth, natural dialogue
• Conversation Flow Monitor: Tracks and adapts the pacing of exchanges
• Dialogue State Tracker: Maintains context across turns
• Cognitive Status Tracker: Continuously evaluates user cognitive state using 6 different Linguistic Biomarkers

Core goal: Predict the cognitive status of people with dementia based on their speech and text using Linguistics (extracting subtle cues from grammar, syntax etc..).

• Extract features like altered grammar, anomia (filler‑word usage), pragmatic coherence, prosodic cues, and slurred pronunciation
• Train machine‑learning and deep‑learning models to produce a composite cognitive‑status score

Technology stack:

• Frameworks & Libraries: Python, PyTorch, Hugging Face Transformers, NLTK, Stanford Parser
• Transformer Models: Llama, Gemma, Phi-3, BERT, MobileBERT
• Machine‑Learning Algorithms: Logistic regression, Random Forest, etc.
• Cloud Platforms: Google Cloud Platform (deployment), Azure (auxiliary services)
• Speech Technologies: TTS, STT, ASR integrations

HCI Collaboration: In partnership with Indiana University’s HCI lab, we built an Progressive Web APP that:

• Visualizes each participant’s cognitive‑level metrics
• Offers interactive memory‑enhancement games
• Includes appointment‑setting and reminder features
• Is refined through workshops with caregiver–patient dyads

Next steps: Integrate the complete platform with QT and the Buddy robot for in‑home trials, enabling more natural, autonomous interactions with people living with dementia.

What I've been building (2026)

• Healthcare AI SaaS platform — RAG, LLM agents, full-stack, deployed end-to-end solo in under 4 months
• ProductivityAI — macOS/iOS app with AI-driven task management, habit tracking, and analytics
• AftercareOS — Multi-layered AI automation platform for the aesthetics/med spa industry
• Multi-agent AI systems with guardrails, security, evaluation pipelines, and observability

Want to connect or collaborate? Feel free to reach out or explore more of my work below.