// he/him · founding_engineer
Scaled payment & SIEM platforms to 10M+ users and 2M+ TPS.
New — ask the AI anything about my work ↘
Currently writing — “Back to First Principles”, a series on Agentic AI ↘
Platform Engineering · Distributed Systems
LLM, RAG & Multi-Agent Architecture
// rag_pipeline · active
// 01. about
From designing mechanical systems to architecting payment platforms serving millions of users, my unconventional journey into software engineering has shaped how I approach building resilient and scalable systems.
I started my career as a Mechanical Engineer working on piping design and stress simulations. While using C/C++ for simulation tooling, I discovered my interest in software engineering. What began as weekend experiments building small applications soon became a full transition into backend development.
I prepared intensively, secured All India Rank 34 in the C-DAC entrance exam, and completed my PG Diploma in Advanced Computing from C-DAC ACTS Pune, which launched my career in software engineering.
My journey in fintech began at Volante Technologies, where I worked on enterprise payment infrastructure contributing to the VolPay platform, recognized by the IBS Intelligence Awards and used by financial institutions such as Goldman Sachs.
I later joined TartanHQ as a Founding Software Engineer, working directly with the CEO and CTO to build products from day zero. I helped build the Perks employee benefits platform, onboarded 30K+ concurrent users generating $66K/month revenue, and designed distributed systems and integrations across multiple providers.
Today I focus on designing high-scale backend systems and event-driven platforms handling large volumes of data and financial transactions. Recently, I've also been working on AI-powered developer tooling and production support agents using RAG pipelines and LLM APIs.
Beyond code, I'm a National Level Taekwondo player who has won multiple gold medals and represented Maharashtra State. Music is my other passion - as a guitarist and vocalist, I've been part of an indie-rock band performing at venues like Hard Rock Cafe and Blue Frog. Whether it's sports, music, or software - I love finding new ideas and turning them into reality from scratch.
Mechanical Engineering Graduate
B.Tech from Shivaji University — non-tech to tech journey begins
Tech Transition · AIR 34
CDAC ACTS Pune — pivoted into software engineering
First Tech Role
Volante Technologies — IBS Intelligence Award for VolPay
Founding Engineer
TartanHQ — built Perks from day zero, 30K+ users, $66K/mo
Enterprise Scale
Mastercard — contactless payments, 10M+ users across Africa
Present
Securonix — scaling SIEM platform to 2M+ TPS
National Level Taekwondo Player
Multiple Gold Medals · Represented Maharashtra State
Indie-Rock Band Member
Guitarist & Vocalist · Hard Rock Cafe, Blue Frog
Product Builder at Heart
Love turning ideas into reality from scratch
// agent_runtime
// 02. experience
Securonix · Cybersecurity | SIEM | UEBA Platform
Jan 2026 - Present · Remote
// 03. projects
Engineered a distributed Multi-Service Payment Wallet system using Java Spring Boot, Spring Security, Kafka, and Redis. Improved transaction processing efficiency by 20% (P95 latency: 200ms to 160ms) and reduced SMS notification latency by 30%.
Built a RAG-based document interaction tool using LangChain, Claude AI/OpenAI, and vector databases. Implemented semantic search with context-aware retrieval for natural language PDF querying with multi-document support.
Created Perks from the ground up at TartanHq, successfully bringing in real customers within just 3 months. Implemented HRMS Connect to onboard 30,000 concurrent users.
Built a comprehensive class management platform handling attendance tracking, notes distribution, and video content delivery. Initially hosted on Raspberry Pi with GitHub Actions CI/CD, later migrated to on-premise server with Google Cloud reverse proxy. Maximized open-source tools to minimize costs while maintaining production-grade reliability.
// 04.5 ai_agent_pipeline · live_demo
Picks a P1 ticket from Jira MCP, gathers context from code, logs, DB, S3 & SQS, then synthesises a root cause + fix with gpt-oss-120b via Groq. — running automatically ↓
SCRUM-1234
Payment processing timeout — NullPointerException on checkout
Occurrences
847 in 30 min
Revenue Impact
~$24K revenue
// agent_runtime · stream
waiting for scroll…
// agent_runtime · 8 agents
Jira Cloud · MCP Server
GitHub · diff analysis
Loki CLI · logql
PostgreSQL · metrics
AWS S3 · config bucket
AWS SQS · queue metrics
Vector store · embeddings
Groq · gpt-oss-120b
// architecture
// 04. skills
LLM APIs · RAG Pipelines · Multi-Agent Architecture · LangChain4j · Vector Databases · Platform Engineering · AI-powered Developer Tooling
// 05. testimonials
Recommendations from colleagues and leaders I've had the privilege to work with.
“Pratik has an innate ability to analyze complex systems and identify optimal solutions. His insightful contributions have significantly enhanced the efficiency and effectiveness of our projects. He possesses a keen eye for identifying potential bottlenecks and takes a proactive approach to optimize performance. Furthermore, Pratik is an excellent team player who fosters a positive and inclusive work environment.”
“Pratik is an exceptional backend engineer with deep expertise in Java, Spring, and building scalable, reliable systems. What stands out most about Pratik is not just his technical strength, but his positive attitude and willingness to go the extra mile for the team. He is someone you can always count on - whether it's solving a complex problem, supporting teammates, or ensuring delivery quality.”
“Pratik is a highly skilled developer who consistently exceeded expectations. They are a quick learner, highly motivated, and always willing to take on new challenges. Their work ethic and attention to detail are unmatched, and they consistently produced high-quality work. Pratik was responsible for designing and developing the Batik platform from scratch with excellent problem-solving skills.”
// 06. writing
LLMs hallucinate facts they were never trained on. RAG turns a closed-book exam into an open-book one — chunk documents on semantic boundaries, embed them, retrieve by meaning, and inject the right passages into context before the model ever answers.
The LLM reads a schema, not your function body — so an imprecise tool description silently corrupts multi-step agent workflows. Covers function calling mechanics, tool selection, parallel execution, and MCP as the emerging standard for wiring agents to tools.
Agents differ from chatbots by completing goals through multi-step reasoning and tool use, not single answers. Building one is system design, not prompt engineering: the ReAct loop, LangChain vs. LangGraph, and where prompt engineering actually fits in.
Every LLM hits three walls: a knowledge cutoff, a context window, and no ability to act in the world. Here's exactly why — and how agents were built to close the gap.
How LLMs actually generate text — one token at a time. Covers tokenisation, BPE, the autoregressive loop, softmax, temperature, and why 'आप कैसे हो?' costs 4× more tokens than 'How are you?'
Nobody teaches a machine the rule. You show it examples and it figures the rule out. Here's what's happening inside a neural network — no jargon, just the actual mechanics.
7 years of writing backend code taught me to break problems into logic. Then I hit fraud detection — and realised some problems refuse to stay still. That's what pulled me into AI.
// 07. contact
Open to interesting conversations, collaborations, and opportunities.