LAT 17.38° N LON 78.48° E GPA 9.35 BUILD v2026.04

ARYA VARMMA

AI engineer. Full-stack developer. CS student. Hackathon winner.

// building RAG pipelines, LLM systems, and full-stack AI products

01 / Identity

About

// who's on the other side of the terminal

./avatar.sh --render

Building AI systems that actually work.

I'm a Computer Science undergrad at VNRVJIET with experience building end-to-end AI products. My focus is on RAG pipelines, LLM integration, and backend APIs - connecting AI models to databases and shipping full-stack solutions using FastAPI and Next.js.

Most weeks I'm designing vector search systems, integrating LLMs into applications, or prototyping new features from concept to deployment. I focus on clean code, testable systems, and solving real problems - whether that's legal document retrieval, anomaly detection, or AI-assisted triage.

Computer Science undergraduate with hands-on experience building AI-powered systems, RAG pipelines, and cloud-native applications. Seeking a role to engineer scalable, production-grade software while contributing to innovative, high-impact projects in a collaborative environment.

9.35 Current GPA
2 Internships
4× 1st Place Wins
02 / Stack

Skills

// the instruments I reach for first

Languages
/01
PythonJavaC++ JavaScriptGoSQL
Frameworks
/02
FastAPIFlaskReact.js Next.jsPyTorchPandas NumPyLangChain
Databases
/03
PostgreSQLMongoDBRedis AWS RDSFAISSPinecone
Cloud & DevOps
/04
AWS S3EC2Glue RDSLambdaDocker TerraformGitHub ActionsCI/CD
Data Engineering
/05
Apache SparkAirflowKafka ETL PipelinesData WarehousingAWS Glue
AI / ML
/06
RAG PipelinesLLM Integration EmbeddingsVector Search Fine-tuningPrompt Engineering
Concepts
/07
REST APIsDistributed Systems OOPAgileLinux Software Design Patterns
03 / Work

Projects

// nine systems, each solving a different hard problem

PRJ_01 AI · RAG

NoteQuest

AI reading tracker with an end-to-end RAG pipeline.

Full-stack reading assistant with document chunking, embedding generation, and cosine-similarity semantic retrieval across 500+ page books. Secured REST APIs, AWS S3 ingestion, adaptive quizzes, chatbot, flashcards, and an LLM-generated chapter-summary layer.

Next.jsFastAPIPostgreSQL FAISSRAGAWS S3OpenAI
500+pages / book
~60%review time cut
PRJ_02 Dev Tools

CodePolice

AI developer alignment & analytics for git history.

Developer analytics system that parses git diff output, generates semantic embeddings per change, and indexes them across 100+ commits for architecture-level drift detection before code review. Ships with a dashboard surfacing task-alignment deviations and anomalous commit patterns.

PythonGit APIVS Code Ext FAISSOllamaFastAPI
100+commits indexed
3+active repos
PRJ_03 Enterprise RAG

PRAGYA

Governance-grade enterprise RAG for Company Law.

Citation-first RAG system using hybrid semantic + keyword retrieval with authority-based filtering across 200+ legal documents. SME-validated ingestion with amendment tracking and audit logging, 100% traceable provenance. Redis caching plus query rewriting.

PythonFastAPIHybrid Retrieval FAISSPostgreSQLRedisTypeScript
200+legal docs
~40%latency cut
PRJ_04 Geospatial

Methane Tracker

Satellite emissions monitoring and forecasting.

Processed 50+ high-res satellite .tif raster files into Cloud-Optimized GeoTIFFs with GDAL, stored in S3, cutting tile load time by 50%+. Dockerized Tiler on EC2 serves 30+ Indian mine sites with side-by-side methane comparison and proximity alerts. Time-series forecasting powers early-warning notifications.

PythonGDALAWS S3/EC2 DockerTiler APILeaflet.jsGeoPandas
30+mine sites
50%+faster tiles
PRJ_05 GNN · Crypto

Chain Signal

GNN-based cryptocurrency anomaly & illicit-wallet detection.

GraphSAGE + GAT models over Elliptic and Elliptic++ blockchain graphs (200K+ nodes, 49 timesteps). Focal loss for class imbalance, temporal split, StandardScaler-persisted checkpoints, threshold calibration capped at 0.70. Ships novelty scoring via illicit-centroid distance, temporal-trend deltas, and a Next.js canvas dashboard with percentile-rank heatmap.

PyTorch GeometricGraphSAGEGAT Focal LossNext.jsFastAPI
200K+nodes indexed
ROC 0.73elliptic test
PRJ_06 AI · Triage

JUSTra

AI-assisted complaint triage for workplace harassment cases.

Auth-free victim intake with OTP gate, unique JUST-XXXX case IDs, and a guided triage chat powered by Groq LLM (llama-3.3-70b). NLP triad scoring decides follow-ups, Groq vision models summarize evidence uploads, and Supabase persists normalized records with severity + urgency for HR and compliance dashboards.

Next.jsSupabaseGroq LLM Vision OCRRAGTelegram Bot
3personas served
JSON-strictLLM contract
PRJ_07 Multimodal RAG

EduRAG

Multimodal RAG for timestamped lecture-video retrieval.

Research pipeline combining Faster-Whisper transcription, SSIM-based keyframe sampling, OCR + adaptive VLM analysis, and a hybrid ChromaDB + NetworkX concept graph. Intent-aware router handles conceptual, temporal, visual, and locator queries. +20% contextual precision over vector-only baselines on 75 QA pairs across three domains.

Faster-WhisperChromaDBVLM NetworkXOllamaGemini
+20%precision vs baseline
75QA eval pairs
PRJ_08 NLP · Verify

Fake News Detector

Claim verification via Telegram bot and web interface.

Designathon 2025 winner. Users submit claims through a React/Vite web UI or Telegram bot; FastAPI backend scrapes source URLs with Trafilatura, processes text with NLTK, and queries Google Custom Search + Gemini to return a verdict with supporting evidence.

ReactTypeScriptFastAPI TrafilaturaNLTKGemini
2surfaces: web + TG
1stDesignathon 2025
PRJ_09 Product

HabitQuest

Gamified habit tracker with XP, streaks, and achievements.

Build habits that stick: daily check-ins mark completion, streaks compound, XP and badges unlock as users level up. Smart reminders, visual progress charts, community sharing, mobile-responsive. One-click completion keeps friction low.

ReactTailwindNode.js MongoDBPush API
XP / Levelsgamified loop
Responsiveall devices
04 / Log

Experience

// recent runtime history, commits to my career

2026.01 / 2026.03 · COMPLETED

AI Engineer Intern

IM India Pvt. Ltd.
  • Engineered an AI-powered legal research platform for Company Law using RAG architecture with hybrid semantic search, enabling contextual retrieval with source attribution.
  • Built a full-stack legal tech application (Next.js, TypeScript, FastAPI, PostgreSQL, Ollama) with document chunking, embedding generation, query rewriting, and Redis-based ingestion tracking.
RAGNext.jsFastAPIOllama
2025.06 / 2025.07 · COMPLETED

Developer Intern

RevGain.ai
  • Worked on production dashboard features including filter optimization and low-code analytics using RgStudio (Next.js, FastAPI).
  • Explored Wingman's AI engine for intent-based business insights, gaining exposure to customer lifecycle analytics, data integration, and cross-team workflows.
Next.jsFastAPIAnalytics
2023 / 2027 · ONGOING

B.Tech, Computer Science & Engineering

VNRVJIET · GPA 9.35 (6th Sem)
  • Active in VNRVJIET Convergence Technical Fest (2023, 2024, 2025) and Convergence 24-Hour Hackathon.
  • Coursework focus: Distributed Systems, Data Structures, Machine Learning, Software Engineering.
GPA 9.35CS & E
05 / Wins

Achievements

// four first-place finishes, and counting

1st place winner VNRVJIET Designathon 2026, gender diversity track ◆ 1ST PLACE 2026
GENDER DIVERSITY

VNRVJIET Designathon 2026

Won 1st place building a solution addressing gender-diversity challenges through product design.

FEB 2026TEAM LEAD
1st place winner Webathon 4.0, full-stack 24-hour sprint ◆ 1ST PLACE 2026
FULL-STACK

Webathon 4.0

Took home 1st in a 24-hour full-stack build sprint, shipping an end-to-end web product on the clock.

FEB 21, 2026FULL-STACK
1st place winner VNRVJIET Design-a-thon 2025, cybersecurity track ◆ 1ST PLACE 2025
CYBERSECURITY

VNRVJIET Design-a-thon 2025

Placed 1st in the cybersecurity track, designing a defensive tool focused on real user threat models.

2025SECURITY
1st place winner Webathon 3.0, product design and full-stack execution ◆ 1ST PLACE 2025
INNOVATION

Webathon 3.0

Won 1st place at Webathon 3.0 for rapid product design and full-stack execution against a tight spec.

APR 05–06, 2025FULL-STACK
/ certifications
05 ISSUED
Machine Learning SpecialisationDeepLearning.AI · Coursera
Data Engineering ProfessionalDeepLearning.AI · Coursera
PyTorch SpecialisationDeepLearning.AI · Coursera
UX Design ProfessionalGoogle · Coursera
Google Cloud Computing FoundationsGoogle Cloud · Coursera
06 / Handshake

Contact

> open to full-time SWE / AI Engineer roles · 2027 grad · based in Hyderabad · open to relocation

View Resume