Generative AI · Cloud · AWS
Tejas PawarAI Systems Engineer
Machine Learning Engineer · Cloud Specialist
Engineering Production-Ready AI Systems from Model to Deployment
Overview
About
Profile Overview
I'm an AI Systems Engineer who builds end-to-end products — modern frontend experiences, backend APIs, and AI-enabled services. I focus on production-ready architectures that are measurable, reliable, and deployable.
Core strengths
- Full-stack engineering (frontend + backend)
- AI systems (RAG, agent-style tooling, evaluation/guardrails)
- Data systems (SQL + graphs + vectors)
Tech I work with
- Backend: Python, FastAPI, REST
- Data: SQL, PostgreSQL, MySQL, Neo4j, vector search
- AI: LLM integration, embeddings, RAG pipelines, prompt guardrails
- Cloud: AWS (Bedrock, IAM, RDS), Docker, deployment workflows
What I'm looking for
AI Engineer / ML Engineer / Applied AI roles (Full-time, Remote/Hybrid)
Location: Newark, DE (ET)
Flagship Case Study
Fitness Workout Tracker & AI Coach
- Full-stack app: workout logging, AI coach (Claude/Bedrock), behavior insights
- Offline-first, cross-platform (web + iOS + Android via Expo)
- Production deployment on Railway — live for recruiters to try
Stack used in this project
Key capabilities
- Workout logging & training plans
- AI coach with daily message + chat
- Offline sync & production auth
Core engineering strengths for production AI systems
Skills
AI Systems Engineering
- RAG & embeddings
- Vector search (pgvector)
- Prompt guardrails & grounding
- Tool / function calling
- Agent-style workflows
- LLM evaluation & failure handling
- Model integration (APIs, cloud)
- Responsible AI design
AI Development Tooling
- Cursor
- GitHub Copilot
- Claude (Projects / Agents)
- OpenAI API
- Prompt Engineering
Data & Backend
- SQL, PostgreSQL, MySQL
- Neo4j (graph modeling)
- Schema design for AI
- FastAPI, Python, REST
- Pydantic, auth, error handling
- ETL & data pipelines
Cloud
- AWS (Bedrock, RDS, S3, IAM)
- Docker & Docker Compose
- Railway
- CI/CD (GitHub Actions)
Case Study
Featured Project
Fitness Workout Tracker & AI Coach
Full-stack fitness app with workout logging, behavior insights, and an AI coach—shipped for web and mobile.
A production-ready fitness app where users log workouts, follow training plans, and get daily guidance from an AI coach powered by AWS Bedrock. The app works offline, syncs when back online, and is deployed on Railway for web and mobile so recruiters can try it live.
Problem
People want to track workouts, see progress, and get simple coaching without switching between multiple tools or losing data when offline.
Solution
A single app that combines workout logging, an AI coach (daily message and chat powered by Claude via AWS Bedrock using real user data), behavior insights (consistency, momentum, dropout/burnout risk, weekly focus), training plans with weekly adjustments and transformation timeline, and offline support. Built and deployed so it's live and usable—recruiters can sign up, try the demo, and experience the full flow.
- Workout logging – Start sessions, add exercises, log sets (reps, weight, RPE), reorder exercises, finish with summary.
- AI coach – Daily message and chat powered by Claude (AWS Bedrock), using the user's real data (metrics, plan, history).
- Behavior insights – Consistency, momentum, dropout/burnout risk, weekly focus derived from workout history.
- Training plan – Set goals (days/week, session length), weekly adjustments, transformation timeline with predictions.
- Offline support – Log and edit workouts without internet; changes sync automatically when connected.
What it does
Workout logging
Start/continue/finish workouts, add exercises from a searchable library, log sets with reps, weight, RPE, and set type; reorder exercises; view previous performance per exercise.
AI coach
Daily personalized message and multi-turn chat (Claude via AWS Bedrock); coach uses only real user data (metrics, plan, history).
Behavior metrics
Consistency score, momentum trend, dropout/burnout risk, adherence type, and weekly focus; computed from workout history and timezone.
Training plan
Set target days per week and session length; weekly adjustments; transformation timeline with body-weight predictions.
Weekly reports
AI-generated narrative summaries of the past week's training.
Auth & profile
Register, login, JWT + refresh tokens; profile with email verification; demo/try mode for quick testing.
Offline support
Read cache and offline queue so users can log workouts without connectivity; sync when back online.
Cross-platform
One codebase for web, iOS, and Android (Expo + React Native Web).
Production deployment
Backend and frontend on Railway; PostgreSQL; health checks; CORS and env-based API URL for production.
Built with
Python · FastAPI · SQLAlchemy 2 · PostgreSQL · Pydantic · Alembic · JWT · bcrypt · AWS Bedrock (Claude) · boto3
React Native · Expo · TypeScript · Zustand · React Navigation · Axios · React Native Web
Railway (hosting) · PostgreSQL (DB) · AWS (Bedrock) · pytest · Jest
Why it stands out
- Live, production app – Not a tutorial clone; deployed so anyone can sign up and use it end-to-end.
- Full-stack ownership – Backend API, frontend app, auth, database design, and deployment pipeline.
- AI integration – Real LLM integration (AWS Bedrock) with structured prompts, context, and error handling.
- Cross-platform – Single codebase for web and mobile; responsive UI and platform-specific details.
- Offline-first – Caching and mutation queue so the app remains useful without connectivity.
- Production practices – JWT + refresh flow, env-based config, CORS, health checks, and clear API versioning (/api/v1).
Sign up or use 'Try Demo' to explore the app. Best experienced on web or in the Expo Go app.
Timeline
Experience
Graduate Assistant (IT-ATS)
June 2025 – May 2026University of Delaware, Academic Technology Services
Newark, Delaware, United States
Contributed to UD Study AiDE (AI-driven learning tool on AWS Bedrock), PATHS Engine (Neo4j knowledge graph), and ProfAI. Built RAG pipelines, backend services, and AI tooling with focus on explainability, retrieval quality, and ethical AI in higher education.
- UD Study AiDE: AI-driven study tool using foundation models on AWS Bedrock; faculty-reviewed, contextually rich study objects; responsible AI design.
- PATHS Engine: Structured knowledge graph (Neo4j) for academic content; retrieval, curriculum mapping, AI-assisted educational tooling.
- Backend services and AI pipelines with Neo4j; data pipelines feeding graph data into retrieval and learning workflows.
- RAG pipelines: ingestion, chunking, embeddings, vector store, retrieval logic; grounded responses; integrated with backend endpoints.
Machine Learning Engineer Intern
March 2023 – July 2023LogicMo Systems Private Limited
Pune, India
Developed and trained CNN models for computer vision (detection, semantic segmentation). Dataset preparation, preprocessing, annotation; architecture design and optimization; training, validation, hyperparameter tuning; data augmentation; integration and deployment documentation.
- Developed and trained CNN models for image detection and semantic segmentation.
- Dataset preparation, preprocessing, and annotation for supervised learning pipelines.
- Designed and optimized neural network architectures; training, validation, hyperparameter tuning.
- Data augmentation for generalization; collaborated on model integration and deployment feasibility.
Credentials
Education
May 2026
M.S. in Data Science
University of Delaware
Professional master's from UD's Graduate College — interdisciplinary program building expertise in data science methods for large-scale and dynamic data, aligned with data engineering and ML systems roles.
Coursework
- Applied Multivariate Statistics
- Applied Multivariate Data Analysis
- Data Mining
- Applied Database Management Systems
- Natural Language Processing
- Mathematical Techniques in Data Science
- Algorithm Design and Analysis
- Unstructured Data Analytics
- Data Computing
- Ethical AI Design
May 2024
B.E. in Electronics and Telecommunication Engineering (Honors in Data Science)
Savitribai Phule Pune University
Premier institution (referred to as the 'Oxford of the East') — B.E. with Honors in Data Science, combining core engineering rigor with applied data science and research-oriented learning.
Coursework
- Machine Learning
- Deep Learning
- Data Structures and Algorithms
- Microcontroller
- Probability and Statistics
- Digital Signal Processing
- Database Management Systems
- Object-Oriented Programming
- Communication Systems
- Computer Networks
What people say
Testimonials
“Tejas is a powerful innovator. In our university department (Academic Technology Services) we habitually design, build, and implement cutting edge solutions that helps bridge the gap between Teaching and Technology. Tejas heard about the impactful Ai project our department was building (UD Study AiDE) and took it upon himself to find if we had need of any assistance. Once we became familiar with not only his skillsets, but his determination, work ethic, and prodigious ability to transform a proof of concept into production-ready solutions and run with it, Tejas became our powerhouse innovator. I personally have overseen many graduate and phd candidates over the 20 years I've worked at the university; Tejas stands out. He is able to effortlessly apply research in practice - which is a fine line many students grapple with. Together we have explored Federated Learning for creating educational AI models, training models for faculty use, and help building out our structured Knowledge Graphs for Ai material generation. Any program would be enhanced by Tejas Pawar's enthusiasm, knowledge, and forward thinking mindset.”
Jevonia Harris
LinkedInUniversity of Delaware, Academic Technology Services
Conference talks, workshops, and professional events
Speaking & Conferences
Speaking at Tech Conference
Live presentation of ProfAI Agent at the Aim Higher East Coast AI Conference 2025 at UD. Demonstrated real-time academic AI assistance and explained the system's architecture and impact.
Conference Panel Discussion
Conference attendees exploring and testing ProfAI during the demo session. Engaged in discussions around AI in education, agent design, and responsible LLM integration.
Slide 1 of 7
Let’s build