I'm Aditya — an IT undergrad (ML minor, class of 2027) building real-time voice agents, multi-user apps, and applied ML pipelines that run in production.
Done — you're booked for Friday, 4:00 PM. I've sent a confirmation by SMS.
~950msPerceived latency
$0/moInfrastructure
~950msVoice agent latency
92.1%MDD detection accuracy
500+DSA problems solved
Top 6%Goldman Sachs Hackathon (16k+)
Selected work
Built end to end, measured honestly.
Voice AI
Voice AI Receptionist
A real-time phone agent that books appointments into PostgreSQL and confirms by SMS on its own — streaming speech-to-text → LLM with function calling → text-to-speech, at ~950ms perceived latency on $0/month infrastructure.
Sticky provider failover and a bounded sliding-window context eliminate dead air and rate-limit failures
Per-turn latency instrumented on a live dashboard
FastAPI service with REST + WebSocket endpoints, deployed via Docker on Render
Per-user isolation at the SQL layer — no query runs unscoped
LeetCode GraphQL API behind a server-side proxy
Session cookies encrypted at rest
CI with auto-migrations on every Vercel deploy
DSA Guardian
A multi-user full-stack app (Next.js App Router) that syncs Striver A2Z sheet progress with your LeetCode profile and recommends company-tagged Hard problems.
An end-to-end EEG pipeline that detects Major Depressive Disorder, with time-, frequency-domain, and nonlinear features validated by cross-validation — packaged as an automated diagnostic system.
Python
Signal processing
ML
Applied ML
Stock Price Predictor
An equity-research pipeline over 2020–2024 daily prices for five large-cap stocks. Compares ARIMA vs XGBoost on time-aware splits (no leakage), scored on RMSE/MAE/MAPE and backtested against buy-and-hold.
Streaming every stage, plus sticky provider failover and a bounded context window, got the voice agent to ~950ms with no dead air — on free-tier infrastructure.
Isolation by default
In DSA Guardian every SQL query is scoped to its user, so no query can run unscoped — not just protected by route guards.
Evaluate honestly
Cross-validation, time-aware splits, and backtests against buy-and-hold, so the numbers I report hold up on data the model hasn't seen.
Experience
Summer Research Intern
IIT Indore · May – Jul 2025
Built an end-to-end EEG pipeline detecting Major Depressive Disorder at 92.1% accuracy, and packaged it into an automated diagnostic system.
Coordinator, E-Cell
SGSITS · Aug 2024 – Present
Led a team of 30+ to run GSPL, an IPL-style campus league fostering entrepreneurship.
B.Tech, Information Technology
SGSITS · Machine Learning minor · 2023 – 2027
Shri Govindram Seksaria Institute of Technology and Science, Indore.
Recognition
Goldman Sachs India Hackathon — top 6% of 16,000+
JEE Advanced — rank 24,989 of ~1.5M candidates
500+ DSA problems solved across platforms
NPTEL — Python for Data Science, Joy of Computing, Intro to ML, Deep Learning, NLP
Stack
Backend & APIs
FastAPI
Node.js
Next.js
REST
WebSockets
JWT / OAuth / RBAC
SQLAlchemy
Pydantic
AI & Data
LLM function calling
XGBoost
Statsmodels
pandas
NumPy
Signal processing
Languages
Python
Java
TypeScript
JavaScript
SQL
C/C++
HTML/CSS
Databases & Tools
PostgreSQL
SQLite
Docker
CI/CD
Git
Postman
pytest
Selenium
Fundamentals
DSA
OOP
DBMS
Operating Systems
Computer Networks
System Design
Have a backend or AI problem worth solving?
I'm open to internships and backend roles. The fastest way to reach me is email.