Priyam Garg
$
Final-year CSE undergrad at LNMIIT Jaipur. I build production AI systems — multi-agent platforms, RAG pipelines, vector search — and the full-stack products that wrap around them. Two SWE internships shipping real infrastructure, and a LeetCode Knight rating from the years of DSA behind it.
➜ whoami
- name
- Priyam Garg
- role
- Software Engineer
- education
- B.Tech CSE · LNMIIT '27
- focus
- AI systems · Full-stack
- stack
- Next.js · Python · LangGraph
- leetcode
- Knight · top 4%
- status
- open to opportunities
➜ cat now.txt
Shipping AI agents that survive contact with real users.
➜
- LeetCode rating
- 1904
- Problems solved
- 680+
- SWE internships
- 2
- Public repos
- 20
$ cat about.md
About
I'm a Computer Science undergrad at LNMIIT Jaipur, graduating in 2027. Most of my time goes into two things: building AI systems that hold up outside a notebook, and staying sharp on the algorithms that make them possible.
AI systems, not AI demos
Multi-agent orchestration with LangGraph, RAG over real corpora, vector stores tuned for sub-second retrieval — and the guardrails that stop them hallucinating in production.
Full-stack by default
Next.js and TypeScript on the front, FastAPI and Node behind it, PostgreSQL/MongoDB and Qdrant underneath. I own features end to end rather than handing them across a wall.
Shipped, not sandboxed
Two internships spent on systems with real consequences — parcel routing for a national logistics network, and a WhatsApp agent serving a thousand-plus live users.
Fundamentals underneath
LeetCode Knight in the top 4% and Codeforces Pupil. Data structures, DBMS, operating systems, and low-level design are the layer everything else sits on.
$ git log --author='priyam' --work
Experience
Two software engineering internships, both spent on systems already carrying real traffic — logistics infrastructure at national scale, and a multi-agent AI platform shipped to a live client.
Xpressbees
Software Engineering Intern
May 2026 – Jul 2026
Pune, India · On-site
Worked on the location intelligence stack behind one of India's largest logistics networks — the systems that decide which hub a parcel routes through and where exactly it gets dropped.
- Built a data migration pipeline to transition core location infrastructure to a new geospatial provider, improving delivery location accuracy across the network.
- Developed an address cleansing pipeline that minimized incorrect address matches and measurably reduced delivery rerouting.
- Built a hub clustering pipeline to fix gaps and overlaps in delivery zones, reducing wrong hub selection.0.5% fewer misroutes
- Built analytics dashboards visualizing delivery-zone clusters and tracking delivery-agent activity, giving ops teams data to act on instead of intuition.
PythonReact.jsSQLData AnalyticsNLPGeospatialKrishna Tulasi Capital LLC
Software Engineering Intern
Jan 2026 – May 2026
Remote · Remote
Built a configurable multi-agent AI platform from the ground up and shipped it to a live client — going from prototype to a production system handling real users on WhatsApp.
- Built a customizable multi-agent AI platform supporting configurable knowledge bases and business-specific workflows.7+ agent types
- Deployed a WhatsApp AI agent for a wellness client, automating onboarding, engagement, and progress tracking end to end.1,000+ users
- Engineered a production-ready system integrating RAG pipelines, contextual multi-turn conversations, and human escalation paths.5+ channels
Next.jsPythonFastAPILangGraphDockerGenerative AIWhatsApp APIVector DBs
$ ls -la ~/projects
Projects
Things I built end to end — architecture, implementation, and the unglamorous parts that make them actually work. Every card links to its source.
Nexus
AI · Full-Stack2026AI-augmented project management platform
Nexus unifies the three surfaces project management normally keeps apart — issue tracking, AI assistance, and the code repository — behind a single retrieval-augmented memory spanning tasks, implementation plans, and source code.
- Architected a platform unifying Kanban boards, roadmaps, analytics, and AI chat over a RAG pipeline that indexes tasks, plans, and code together.
- Engineered a LangGraph agentic workflow (Planner → Critic → Finalizer) for requirement-to-task decomposition, with Zod-validated structured outputs.
- Designed a three-collection Qdrant vector store (tasks, roadmap, code chunks) using 768-dim Gemini embeddings with payload indexes for sub-second context retrieval.
- Wired GitHub webhook sync verified by HMAC-SHA256 signatures, matching commits back to the tasks they close.
- Enabled real-time multi-user board sync via Supabase Realtime and Zustand, with owner/admin/member role-based access control and invite-based team management.
- Shipped an analytics dashboard with burndown charts and completion metrics on top of Drizzle-managed PostgreSQL migrations.
Stack
MOCKVIEW
AI · Full-Stack2026Real-time voice-based mock interview platform
A mock interview platform that actually listens. Three interview modes, AI-generated questioning that adapts to your answers, live speech recognition, and coaching on the things nobody grades you on — communication, pacing, and confidence.
- Built a real-time voice-based interview experience with 3 modes, AI-generated questioning, speech-to-text recognition, and live coaching feedback.
- Implemented resume ATS analysis across 13 target roles — PDF parsing, keyword extraction, a 5-step analysis chain, role-weighted scoring, and actionable rewrite suggestions.
- Designed a dashboard tracking 4 performance dimensions: interview scores, streaks, and progress metrics for continuous improvement.
- Integrated secure authentication via NextAuth and Prisma over PostgreSQL to manage accounts, sessions, and full interview history.
Stack
$ ls ~/projects/more
A legal document Q&A assistant that ingests PDFs and answers questions over them — with a similarity threshold that makes it decline out-of-scope queries instead of hallucinating law.
- RAG pipeline over legal PDFs using LangChain, Gemini Flash 2.0, and multilingual Sentence-Transformer embeddings stored in Qdrant Cloud.
- Similarity-threshold refusal: queries below a tunable relevance floor are declined rather than answered, cutting hallucination on out-of-domain questions.
- Persistent per-user, multi-session chat history in MongoDB, secured with JWT + bcrypt auth.
An education platform that generates full courses and quizzes on demand — a one-stop exam-prep surface built on Groq's Llama 3.3 70B, in both English and Hindi.
- AI-driven course and quiz generation powered by Groq (Llama 3.3 70B), with real-time progress tracking for long-running async generation jobs.
- Seven-model Prisma schema (User, Course, Unit, Chapter, Quiz, QuizAttempt, GenCourse) over PostgreSQL.
- Bilingual English/Hindi content, light and dark themes, and a built-in AI chatbot for learning assistance.
- Multi-provider OAuth via NextAuth.js with bcrypt-hashed credential fallback.
A full-stack agricultural advisor that turns soil nutrients, live weather, and GPS location into a concrete answer: what to plant. Built as an offline-capable PWA in English and Hindi, for the field rather than the desk.
- Scikit-Learn crop recommendation model served from a Python/Flask microservice behind a Node + Express TypeScript API.
- Live weather with 7-day forecasts and 90-day seasonal climate analysis via OpenWeather and Open-Meteo.
- GPS farm mapping with manual entry fallback, JWT auth, and MongoDB persistence.
- Progressive Web App that keeps working offline — the deployment constraint that actually matters for rural users.
An XGBoost pipeline forecasting Bitcoin price movement across 15-minute, 30-minute, and 1-hour timeframes, built to study how feature engineering choices change what a model can actually see.
- End-to-end notebook pipeline: ingestion, cleaning, feature engineering (returns, rolling windows, realized volatility), split, train, evaluate.
- XGBoost classifier and regressor variants trained across three BTC/USD timeframe datasets.
- Evaluation via accuracy metrics and confusion matrices on held-out buy-signal labels.
$ cat package.json | jq .skills
Technical skills
The tools I reach for, grouped by where they sit in the stack — plus a live breakdown of what my public repositories are actually written in.
Languages
- C
- C++
- Java
- JavaScript
- TypeScript
- Python
- SQL
- HTML
- CSS
Backend & Data
- Node.js
- Express.js
- FastAPI
- REST APIs
- WebSockets
- Webhooks
- MongoDB
- PostgreSQL
- Vector Databases
Frontend
- React.js
- Next.js
- TailwindCSS
- ShadCN
- Zod
- NextAuth
- Framer Motion
AI / ML
- LangChain
- LangGraph
- RAG
- Generative AI
- Vector Search
- Embeddings
- Prompt Engineering
- Scikit-learn
- Pandas
- NLP
Tools
- Git
- GitHub
- Docker
- Postman
- Supabase
- Figma
- Jira
- AppScript
- Linux
CS Fundamentals
- Data Structures & Algorithms
- OOPs
- DBMS
- Operating Systems
- Low-Level Design
- Computer Networks
Languages across my repositories
live$ ./solve --daily
Problem solving
The part that doesn't show up in a repo. Consistent competitive programming is why I can reason about complexity before I write the first line.
LeetCode
@priyamgarg1904
Contest rating
Knight
Badge · top 4.48%
18
Contests
Codeforces
@priyamgarg8ii1371
Max rating
Pupil
Peak rank
1371
Current
Rated contest participation on Codeforces — timed problem solving under constraints that practice sets don't reproduce.
View profileTopics I work in
$ cat education.json && cat awards.json
Education & achievements
Education
The LNM Institute of Information Technology (LNMIIT)
Aug 2023 – May 2027B.Tech, Computer Science & Engineering
CGPA 7.21 / 10· Jaipur, Rajasthan
Cambridge Court World School
2023Class XII — CBSE
91 / 100
Neerja Modi School
2021Class X — CBSE
94.66 / 100· Jaipur
Achievements & leadership
LeetCode Knight
Top 4% globally · contest rating 1904.
Codeforces Pupil
Max rating 1371.
Elected Senator
Elected to the LNMIIT student body senate by 600+ voters.
$ ./contact --priyam
Let's build something
I'm open to software engineering and AI engineering roles, and always up for a conversation about agents, retrieval systems, or a hard problem you're stuck on.
Jaipur, Rajasthan · India