Aditya Sidham
Résumé
Open my résumé

I'm Aditya Sidham. I build AI that sees.

Applied AI, iOS and backend engineer in Los Angeles. M.S. Applied Data Science at USC.

Demos are easy. I build the version people use.

A patent-pending form coach. A four-agent nutrition app. A database you can ask in plain English.

One person. The whole stack.

SwiftUI on the phone, FastAPI and Postgres behind it, Claude agents doing the reasoning. I built every layer of FitandFine myself.

Find the points that matter.

A pose model keeps 33 points from a noisy frame. I treat messy problems the same way: find what matters, then build on it.

Let's build something that holds up.

Open to full-time roles in AI, iOS and backend, and to early-stage teams building something real.

Start a conversation See the work

I'm Aditya Sidham. I build AI that sees.

Applied AI, iOS and backend engineer in Los Angeles. I take products past the demo: a patent-pending form coach, a four-agent nutrition app, a database you can ask in plain English.

Start a conversation See the work
Skip to the work

Things I've shipped.

Each one started as a practical question. Can a camera coach your form? Can someone query a database without knowing SQL? Can a nutrition app adjust to what you actually eat? I build the whole loop, then fix what breaks when real people use it.

FitandFine dashboard: calorie ring at 0 of 2000 kcal, protein, carb and fat bars, and a weekly streak row FitandFine food log grouped into breakfast, lunch, dinner and snack with per-item macro chips FitCoach answering a protein question using the user's own logged data FitandFine analytics: 30-day averages, calorie trend bars against goal, and a macro split chart

Flagship, 2026, solo build

FitandFine

A nutrition app that learns from what you actually do. I built all of it: the SwiftUI app, the FastAPI backend, the label OCR pipeline, and four Claude agents that turn a label photo into a plan you can push back on.

Barcode to food card
<500ms
USDA foods indexed
50k+
Label OCR field accuracy
85%+
Claude agents
4

The hard part: text read off a label photo is noisy. The parser scores every field, and anything it is unsure about goes to a review screen instead of straight into your log.

Read the case study
  • SwiftUI
  • SwiftData
  • HealthKit
  • FastAPI
  • PostgreSQL
  • Redis
  • Celery
  • Claude API
  • AWS
FitForm counting dumbbell reps with a pose skeleton overlaid on the webcam feed patent_app 202421099595

FitForm

A camera that counts a rep only when the form is right. It tracks 33 body landmarks at 25 FPS on a laptop CPU, corrects you on screen mid-set, and led to a published patent application.

Computer vision, 2024

  • Python
  • MediaPipe
  • OpenCV
  • TensorFlow
Chat-DB turning an English question into a MySQL query query_time −60%

Chat-DB

Ask a database a question in plain English and get a query you can check before it runs. Query writing got about 60% faster, and answers were 92% correct on a 150-question test suite.

Team lead of four, USC, 2025

  • Python
  • Streamlit
  • Gemini API
  • MySQL
  • MongoDB

Where it started

Aditya Sidham standing on a trail above Los Angeles, the city and the San Gabriel mountains behind him
Press and hold to run pose detection on this photo. With a keyboard, hold Space or Enter.

I learn by getting close to the problem.

  • NowM.S. Applied Data Science, USC, finishing December 2026
  • BeforeB.Tech Computer Engineering, Data Science specialization, DY Patil RAIT
  • Based inLos Angeles, California
  • FromMumbai, India
  • I build withAI, iOS, backend systems, data, computer vision, hardware
  • Off screenStrength training, calisthenics, running, long walks that turn into project ideas

I've never been satisfied knowing only the part of a system assigned to me. If I'm using a tool, a sensor or a model, I want to know what it's doing, where it fails and how it fits into everything around it. That instinct has taken me through Arduino circuits, computer vision, enterprise backends, native iOS and multi-agent AI.

I started in physical computing, debugging with a multimeter as often as a terminal. Computer engineering gave me the software foundations. At TCS and LTM my code sat inside larger operational workflows, so correctness and reliability mattered. At USC I've gone deeper into machine learning, data mining and language systems.

FitandFine brings those threads together. I own the interface, the backend, the data model, the OCR pipeline and the AI layer at once, which forces me to think about the user and the architecture at the same time. That's the work I want to keep doing.

I build to understand

Documentation is a starting point. A working prototype tells me what the real problem is, so I get something running early and improve it with evidence.

I think in complete systems

A model only helps if the data reaches it reliably and the result makes sense to a person. I care about the whole path from input to interface.

I like the unglamorous parts

Caching, validation, error states, background jobs and offline behavior decide whether a demo becomes something people can depend on.

I ask questions early

Owning a problem doesn't mean going quiet. I check assumptions, share progress and raise risks before they get expensive.

Applied AI and LLMs

  • Claude API
  • Gemini API
  • Multi-agent orchestration
  • RAG
  • MCP
  • LLM evaluation
  • Claude Code

Vision and ML

  • OpenCV
  • MediaPipe
  • PyTorch
  • TensorFlow
  • Scikit-learn
  • Pandas and NumPy

Languages

  • Python
  • Swift
  • Java
  • C#
  • TypeScript
  • JavaScript
  • C++
  • SQL
  • Bash

Apps and frontend

  • SwiftUI
  • SwiftData
  • HealthKit
  • AVFoundation
  • React
  • Streamlit

Backend

  • FastAPI
  • Spring Boot
  • ASP.NET
  • REST and microservices
  • Celery
  • SQLAlchemy

Data

  • PostgreSQL
  • MySQL
  • SQL Server
  • MongoDB
  • Redis
  • Apache Spark
  • ETL pipelines

Cloud and DevOps

  • AWS EC2
  • AWS S3
  • AWS Lambda
  • API Gateway
  • CloudWatch
  • Docker
  • Linux
  • Git
  • CI/CD

Testing

  • Unit and integration tests
  • pytest
  • Code reviews
  • Design patterns

Hardware

  • Arduino
  • Raspberry Pi
  • ESP8266
  • MQTT
  • Servo systems
  • Sensor integration

Goals you have to earn, one session at a time.

Training made me patient with slow progress and honest about feedback. You can't talk your way around a missed workout or a weak system. You put in the work, measure what happened, and adjust. It's the same loop I run on code.

  • 5KTempo
  • 10KThreshold
  • 21.1KHalf marathon
  • 30KLong run
Strava Follow the training log
latest_activity

Where I've done the work.

Large delivery teams and small hands-on projects. Different places, same expectation: understand the system, make the change carefully, leave it working better than before.

  1. Jan 2025 – Dec 2026 Now

    M.S. in Applied Data Science

    University of Southern California, Los Angeles. GPA 3.52.

    • Coursework in natural language processing, machine learning for data science and data mining.
    • Led the four-person Chat-DB team. Building FitandFine alongside the degree.

  2. Sep 2024 – Dec 2024

    Software Engineer

    LTM, formerly LTIMindtree, Mumbai.

    • Built C# and ASP.NET backend integrations for Otis's enterprise application, across service management, dispatch, inventory and reporting.
    • Rebuilt the SQL Server procedures behind IoT fault-event ingestion and work-order creation, cutting backend issue resolution time by 20%.

  3. Jan 2024 – May 2024

    Software Development Intern

    Tata Consultancy Services, Navi Mumbai.

    • Built a full-stack module for CPMS, the Central Plan Scheme Monitoring project, in Java, Spring Boot, JSP, Servlets and SQL.
    • Automated repeated admin workflows, cutting manual data entry time by 35%, shipped in two-week sprints.

  4. Jun 2020 – Jun 2024

    B.Tech in Computer Engineering, Data Science specialization

    DY Patil Ramrao Adik Institute of Technology, Navi Mumbai. GPA 3.68.

    • Projects across computer vision, IoT and robotics. Led the four-person SafeSense team to the CIIA finals.
    • Built FitForm, which later became a published patent application with me as a co-applicant.

  5. May 2023 – Jun 2023

    IoT Intern

    Aces Invensys Technologies, Navi Mumbai.

    • Built a Python and MQTT pipeline that turned live IoT beacon signals into location estimates.
    • Tracked beacons to within about ±10 cm in testing, working through noisy sensor data and calibration.

Landmarks so far.

A published patent application, a solo-built AI product, a national finals and a maker award, plus two programs that taught me to tie technical choices to business context.

Jan 2025

Patent application: FitForm

Co-applicant on Indian patent application 202421099595, "AI Powered Exercise Tracking and Form Correction System", published 17 January 2025.

View the case study

2026

FitandFine, solo-built

Label OCR to structured macros at over 85% field accuracy, weekly diet analysis, meal suggestions and a calorie target that adjusts to real data.

View the case study

2023

CIIA finalist

Led a team of four that built SafeSense, selected as a Creative Ideas & Innovations in Action finalist.

View the case study

Sep 2018

SP Robotics Maker of the Month

For a servo-driven robotic arm I designed and programmed, later extended with gyroscope gesture control.

View the case study

Jun 2024

Google Data Analytics Professional Certificate

Data preparation, analysis, visualization and communicating findings, completed on Coursera.

Jun 2023

Harvard Business School Online

Business Analytics and Sustainable Business Strategy. Both taught me to connect technical decisions to evidence and long-term trade-offs.

Have a hard problem? I'd like to hear it.

I'm looking for full-time roles in AI, iOS and backend engineering, and for early-stage teams that need someone to own a problem end to end. If you're building something that has to work past the demo, tell me what you're trying to solve.

sidham@usc.edu