Namaste! I’m

App mode: ON, still me,

Pranav Deshpande

I build whole products, end-to-end.

I ship mobile apps people actually use.

Software engineer. I like owning the whole path from schema to shipped: the data model, the API, the interface, and the release pipeline behind it. Python, Java, C++ and TypeScript, picked per problem rather than per habit. B.Tech in AI & ML at Manipal University Jaipur, class of ’27.

Mobile engineer. I build apps in React Native and Expo, wire them to APIs I wrote myself, add the AI features when they earn their place, and own the release pipeline that gets them past App Store review. B.Tech in AI & ML at Manipal University Jaipur, class of ’27.

GitHub LinkedIn

the shell works — run one!

pranav@portfolio: ~ demo shell

pranavsh 1.0, a hand-built demo shell with no real process behind it. Pranav builds software end-to-end: mobile, backend, infra, and the AI when the product needs it. Run something ↓

whoami ls ~/projects cat stack.txt ./hire.sh

the phone works — tap an app!

Now shipping: Sprout SDE Intern · SeedlingLabs B.Tech AI & ML MUJ · CGPA 8.63
Python Java C++ TypeScript SQL FastAPI PostgreSQL Redis Celery Docker AWS GitHub Actions React React Native PyTorch LangGraph RAG pipelines

In production

Real apps, real users, real deadlines: the work that backs the claim.

SDE Intern · June 2025 — present

Sprout — SeedlingLabs

Sprout is SeedlingLabs’ D2C mobile app; my job is building it. I orchestrate the whole thing end-to-end, frontend architecture and backend delivery, and I own what happens after the code is written too: monitoring, migrations, releases, and App Store review.

  • Designed and built an AI Tutor feature end-to-end: prompt architecture plus a student-facing AI dashboard, bringing GenAI into the core product
  • Raised reliability with Sentry crash and error monitoring: resolved N+1 API calls and UI flickering, fixed backend DB migration failures
  • Built the CI/CD and OTA release pipelines for the D2C and B2B apps, streamlining deployments and App Store review/compliance fixes
Sprout · UI abstracted

Trainee Software Developer Intern · June — July 2025

AI intrusion detection — Ithena Technologies, Pune

Built a real-time AI intrusion detection system with YOLO and IoT: live alerting, with inference tuned for low latency. Designed the MQTT/ThingsBoard telemetry pipeline and built the custom dashboards and mobile interfaces for analytics.

The toolbox

Backend, cloud, frontend, mobile, AI. I pick the tool the problem needs, not the one I already know.

Languages

  • Python
  • Java
  • C++
  • JavaScript / TypeScript
  • SQL

Backend & data

  • FastAPI
  • PostgreSQL
  • Redis
  • SQLAlchemy
  • Alembic
  • Celery
  • RabbitMQ

Cloud & DevOps

  • Docker
  • AWS
  • GitHub Actions
  • Prometheus
  • Grafana
  • Sentry
  • Unleash

Frontend & mobile

  • React
  • React Native
  • Expo

AI / ML

  • PyTorch
  • HuggingFace Transformers
  • OpenCV
  • scikit-learn
  • Pandas
  • NumPy
  • Computer Vision
  • NLP

GenAI & LLMOps

  • RAG pipelines
  • Qdrant
  • PGVector
  • Embedding models
  • Semantic search
  • Prompt engineering

Things I’ve built

Three systems, three deliberately different stacks: a code-execution trace engine, a payments layer agents can safely call, and a database agent. Click one for the full story.

VisionDS: a DSA visualiser that traces your own code

0 servers student code never leaves the browser

Paste a Python solution and its testcases, watch an animated dry-run of the code you actually wrote, and jump straight to the step where your output diverges from the expected one. Everything executes in your own browser.

React · TypeScript · Vite · Zod · Pyodide (WASM) · Web Workers · pnpm monorepo

Read the write-up Code ↗

Latch: an agentic commerce layer for appointments

7 tools any agent can call, no partnership needed

A service-transaction layer on a merchant's Razorpay account, exposed as an MCP server with seven tools, so any third-party agent can hold a slot, read the cancellation ladder, pay a deposit and reschedule. Deployed and running against real Razorpay test mode.

TypeScript · Node.js · Fastify · PostgreSQL · Drizzle · MCP (Streamable HTTP) · Razorpay API · Zod · Vitest · Railway

Read the write-up Code ↗

Natural language to SQL agent

2 agents routed per turn, SQL or conversation

A session-based, RAG-powered LangGraph agent that turns plain English into SQL and runs it against a live PostgreSQL database, routing between a SQL agent and a conversational one, retrieving few-shot examples by vector search, and holding risky queries behind human approval.

LangGraph · LangChain · OpenAI · PGVector · PostgreSQL · SQLAlchemy · FastAPI · Docker · LangSmith · Python (uv, Ruff)

Read the write-up Code ↗

more shipping soon — this space is reserved ✦

More on GitHub: repos, contributions, and the code behind all of this.

Beyond the repo

Leading

Chairperson of the MUJ ACM SIGAI student chapter, driving AI-focused initiatives and collaborative projects for a 300+ member community.

Learning

B.Tech in Artificial Intelligence & Machine Learning at Manipal University Jaipur, 2023–2027. CGPA 8.63. Coursework in DSA, DBMS, machine learning, deep learning, NLP and GenAI.

Practicing

Currently going deep on system design and DSA, with daily problem-solving reps:

LeetCode ↗

Certified through Cisco NetAcad (Python, Networking), NPTEL (DSA, DAA) and Coursera (Agile).

open to opportunities

Let’s talk.

Looking for an engineer who can own the whole thing: app, backend, and the AI inside it? My inbox is open. (I reply fast!)

pranavdeshpande2005@gmail.com

LinkedIn GitHub