Architecting production AI systems, modern full-stack web applications, and verified backend pipelines.
From responsive React & Next.js interfaces with MongoDB/PostgreSQL backends to multi-language AST transpilation and SLURM distributed pipelines. I build dependable systems that operate reliably at scale.
Test my system architecture live
Select an architecture below to see how I design, automate, and stress-test end-to-end AI pipelines for real client production.
Evaluated across 9 languages. Pairs AST semantic mapping with LLM synthesis and automated compiler-feedback retry loops (gcc, javac, tsc).
Real-Time Pipeline Execution Flow
Looking for a custom version of this pipeline?
I can build, test, and ship this directly to your cloud, Kubernetes cluster, or server.
Selected work
A few projects worth a closer look
Each one started as an ambiguous problem. Here's what shipped, and what changed because of it.
BigMath — Verified AI Training Data Pipeline
Designed and orchestrated an end-to-end pipeline processing 100K+ mathematical problems across 10+ datasets (DeepMath, LIMO, AIME, TheoremQA, etc.). Covers MCQ conversion, Evol-Instruct scaling with 5 mutation strategies, teacher reasoning generation, symbolic verification, and difficulty tagging — producing 53K+ verified reasoning traces.
Multilingual Harmful Meme Detection & Web App
Curated and annotated 13,700+ multilingual memes, creating the first open-source benchmark datasets for Telugu and Kannada harmful meme detection. Implemented a MOMENTA-inspired multimodal pipeline integrating IndicBERT, CLIP, VGG-19, and ConceptNet, paired with a modern responsive web interface for live interactive inference.
AI Diagnostic Report Generator
Built a document-intelligence system that extracts text and images from inspection and thermal PDFs, then uses the Claude API to assemble a seven-section diagnostic report in structured JSON — explicitly flagging missing or ambiguous data.
LLM Code Transpilation Pipeline
Architected a repository-level transpilation pipeline using a two-queue threaded design integrating AST dependency analysis, real compiler validation (gcc, javac, tsc), and an automated compiler-feedback retry loop. Scaled to process 1.8M+ files and 6.1B tokens across 72 language pairs.
ACID-Compliant B+ Tree Database Engine
Designed a B+ Tree storage engine from scratch with O(log n) operations, full ACID transaction support via Write-Ahead Logging, shared/exclusive lock-based concurrency control, ARIES-inspired crash recovery, and a FastAPI layer exposing 40+ REST endpoints.
AutoStream Social-to-Lead Agent
Built an AI-powered lead-generation agent that monitors social media signals, classifies purchase intent, and converts engagement into structured lead data for sales pipelines.
Services & Capabilities
What I build for clients
Specialized engineering across modern full-stack web applications, applied AI systems, and high-performance backend infrastructure. Built for reliability, precision, and production scale.
Full-Stack Web Applications
I architect end-to-end full-stack web applications — from pixel-perfect, responsive UI design to database schemas and robust backend APIs. Focused on high-performance Core Web Vitals, seamless UX, and clean, type-safe architecture that scales effortlessly.
AI Agents & LLM Applications
I design and build LLM-driven applications — from intelligent document processors to multi-step reasoning pipelines. Whether you need a RAG system, an AI agent, or a domain-specific chatbot, I architect it with reliable validation, structured outputs, and real compiler or rule-based verification.
Backend APIs & Microservices
I build FastAPI and Node.js backends with 40+ endpoints, full test suites, and proper SQL/NoSQL integration. Whether you need a standalone microservice or a high-throughput data-ingestion API, I deliver well-documented, type-safe code your team can extend.
Machine Learning Solutions
From multimodal classification systems to fine-tuned transformer models, I build ML solutions grounded in proper experimental methodology — ablation studies, metric tracking, and clear evaluation on held-out data.
Python Automation & Custom Scripts
Need a script that processes thousands of files reliably overnight? I build batch-processing systems with checkpointing, retries, and error recovery so your workflows don't silently fail. Tested with pytest, documented, and maintainable.
Intelligent Document Processing & RAG
I build document-intelligence systems that extract text and images from complex PDFs, apply LLM reasoning, and produce structured, auditable outputs. Built-in missing-data flagging means you always know what the AI found — and what it didn't.
Have a custom AI or Python requirement not listed here?
I frequently build bespoke pipelines, scraping engines, and customized APIs.
Process & Reliability
How I work with freelance clients
A transparent, milestone-driven workflow designed to deliver production-ready software without surprises or delays.
Understand Your Requirements
Phase 01 of 05I start by listening carefully. I ask detailed questions to understand the problem you're solving, the constraints you're working with, and what success looks like for your project.
Plan the Technical Solution
Phase 02 of 05Before writing a line of code, I map out the architecture, select appropriate tools, and share a clear plan. No surprises mid-project.
Build Iteratively
Phase 03 of 05I develop in stages with regular updates. You see progress early and can give feedback throughout — not just at the end.
Test & Validate
Phase 04 of 05Every deliverable is tested against real inputs. For AI systems, I validate on edge cases and document known limitations honestly.
Deliver & Document
Phase 05 of 05Clean, commented code with a README and setup instructions. You'll know how to run, maintain, and extend what I build.
Regular milestone summaries, Loom walkthroughs, and clear Slack / WhatsApp communication.
Every deliverable includes automated pytest test suites and schema validation for edge cases.
Complete setup guide, Dockerfiles, and environment configs so your team can deploy immediately.
Track Record & Education
Where I've engineered systems
Real production experience at high scale: multi-language compiler validation, 1.8M+ file batching, and formal AI education.
Technical Staff Intern
atSoket AI LabsDistributed Systems & LLM EngineeringHigh-throughput compiler validation and distributed LLM transpilation pipelines
- ▹Evaluated deterministic transpilation across 9 programming languages, benchmarking AST mappers and identifying semantic-gap limitations that motivated a hybrid LLM approach (GPT-O3/Qwen3-30B).
- ▹Architected a repository-level transpilation pipeline using a two-queue threaded design, integrating AST dependency analysis, real compiler validation (gcc, javac, tsc), and automated compiler-feedback retry loops.
- ▹Scaled distributed execution across SLURM and Kubernetes for all 72 language pairs — engineered a sliding-window batch controller (36 pods, 180 concurrent requests) to process 1.8M+ files and 6.1B tokens.
- ▹Diagnosed and resolved critical production bottlenecks including multi-threaded queue race conditions, Kubernetes naming collisions, and LLM request timeouts; implemented file-level checkpointing to ensure zero data loss on pod evictions.
B.Tech. in Artificial Intelligence
atIndian Institute of Technology, GandhinagarComprehensive study of machine learning, system architectures, database engines, and high-performance algorithms
Core academic and laboratory coursework encompassing machine learning algorithms, database internals (B+ trees, WAL, ACID concurrency), computer vision, natural language processing, and scalable distributed computing.
Background & Philosophy
Engineering software that survives production
A background rooted in formal artificial intelligence, distributed systems scale, and relentless attention to edge cases.
I am an AI Systems & Full-Stack Web Developer and undergraduate at IIT Gandhinagar specializing in Artificial Intelligence. My work sits at the intersection where modern web interfaces, reactive frontend architectures, and high-reliability AI backends converge.
During my engineering work at Soket AI Labs, I designed compiler-feedback validation loops and scaled code-transpilation pipelines across Kubernetes and SLURM to process over 1.8 million files and 6.1 billion tokens without data loss.
Alongside distributed AI infrastructure, I build production-grade web applications — combining modern React, Next.js, and TypeScript frontends with MongoDB/PostgreSQL backends, REST APIs, and automated test harnesses. Whether building user-facing web tools, ACID-compliant database engines, or complex LLM pipelines, I engineer software built to withstand real production usage.
When collaborating with founders, engineering leads, and freelance clients worldwide, I prioritize clear asynchronous communication, rigorous test suites, and turnkey codebases that your internal team can deploy and maintain effortlessly.
Deterministic Guardrails First
I never rely on bare LLM prompts for mission-critical logic. Every output is strictly bounded by Pydantic schemas, AST diffing, or automated compiler feedback loops.
High-Throughput Concurrency Native
Having scaled jobs across 36 pods and 180 parallel workers, I engineer sliding-window batching, token-bucket rate limits, and atomic checkpointing into every pipeline.
Turnkey Handover & Pytest Suites
No messy scripts. Every client engagement includes Dockerized environments, 90%+ automated test suites, OpenAPI documentation, and a 30-day post-launch warranty.
Technical Stack & Tooling
Tools and disciplines I build with
Categorized by domain with zero arbitrary percentage bars — each technology backed by real production code, verifiable projects, and accredited coursework.
Full-Stack Web
AI & Machine Learning
Backend & Systems
NLP & Computer Vision
Data & Distributed Systems
Developer Tools & DevOps
Certifications & Accreditations
Formal specializations covering Machine Learning algorithms, full-stack web application development, and distributed systems — each verified via Coursera with permanent certificate IDs.
Supervised Machine Learning: Regression and Classification
Instructor:Andrew Ng (Adjunct Professor, Stanford)
Rigorous foundations of predictive modeling, vectorization, gradient descent mechanics, and multi-class classification from Stanford University and DeepLearning.AI.
Key Competencies
Unsupervised Learning, Recommenders, Reinforcement Learning
Instructor:Andrew Ng (Adjunct Professor, Stanford)
Advanced ML systems covering anomaly detection, collaborative filtering recommenders, PCA, and deep reinforcement learning with continuous and discrete action spaces.
Key Competencies
Web Application Development with JavaScript and MongoDB
Instructor:Dr. Matthew Yee-King
Full-stack web application engineering covering dynamic client-server data flows, document-oriented NoSQL databases with MongoDB, reactive event-driven architecture, and secure session management.
Key Competencies
Responsive Web Design & Modern Frontend Development
Instructor:Dr. Matthew Yee-King, Dr. Kate Devlin, Dr. Marco Gillies
Mastery of modern responsive layouts, fluid typography, mobile-first design patterns, CSS Grid and Flexbox systems, and cross-browser interactive interfaces built with clean JavaScript.
Key Competencies
Computer Communications Specialization (4 Courses)
Instructor:Prof. Xiaobo Zhou
Comprehensive 4-course specialization covering fundamental computer networking, protocol design principles, TCP/IP programming, packet switching algorithms, and network layer architectures.
Key Competencies
Introduction to Meteor.js Full-Stack Development
Instructor:Prof. Matthew Yee-King
Architecting real-time reactive web applications with full-stack JavaScript, covering data synchronisation, reactive templates, and cloud application deployment.
Key Competencies
Lifelong Engineering & Verification
All credentials above are cryptographically verifiable and backed by hands-on project deliverables.
Recognition
Selected achievements
IIT Gandhinagar Explorer Fellowship
Competitive experiential learning programme for real-world AI research, selected competitively from across IIT Gandhinagar.
Dakshana Foundation Scholar
Selected through a rigorous national entrance exam, ranking in the top 1% of candidates nationwide.
Academic Honors & Distinctions
Consistently achieved top academic percentiles in national examinations (Class XII: 92.60%, Class X: 90.00% at J.N.V. Raichur).
Have an AI project or technical challenge? Let's build it together.
Usually replies within 12 hours. Open to milestone-based freelance contracts, bespoke AI systems, and technical consultations worldwide.