AI / Machine Learning
Supervised & unsupervised learning, feature engineering, model evaluation, pipeline design.
Loading…
AI EngineerAI/ML Problem Solver Building Real-World Intelligent Systems
From data to deployment — designing AI products that actually work in production.
Problem solver focused on building end-to-end AI and ML systems that turn real-world problems into scalable, production-ready solutions.
01/ About
AI/ML problem solver building end-to-end systems.
Hi, I’m Parnish (Trilochan) 👋
An AI/ML Engineer & Problem Solver who thrives on building end-to-end intelligent systems — from problem understanding and data analysis to production-ready deployment 🤖⚙️
But before titles, roles, and tech stacks — I’ve always been someone driven by curiosity.
My journey with technology didn’t start with code or AI models.
It started when I was a kid playing with motors, wires, and broken toys 🔌😄 — trying to make fans spin, lights glow, and things move, not because I needed them, but because I wanted to see if I could.
I remember genuinely wondering if there was a person inside the TV controlling everything 📺😂
That curiosity slowly evolved into questions about visuals, signals, electricity, logic — even before I had words for them.
As I grew older, that curiosity didn’t disappear.
It just became more focused.
While others were happy using technology, I wanted to understand what happens underneath 🤔
Why does this button exist?
What happens after I click it?
How does data move from one place to another?
I didn’t always have the best resources.
For a long time, I didn’t even have my own internet connection 😅
I had to borrow access from neighbors, depend on limited availability, and make the most of whatever was possible.
People sometimes mocked it — calling it a waste of time or questioning why I spent so much effort just to stay connected.
But for me, staying connected meant staying curious, learning, and growing.
So I learned how to manage connections, work within limitations, and make things work with what I had 🌐
Not because it was impressive — but because learning mattered more than comfort.
as a great person said “Necessity is the mother of invention.”
At some point, I had to make a choice.
While many people around me focused purely on grades and exams, I chose to focus on skills and systems ⚙️
That path was slower, harder, and full of doubt — but it taught me something important:
Marks look good on paper. Skills build real things.
That belief shaped how I learn and how I work.
I don’t build AI for demos, notebooks, or buzzwords.
I build real, scalable, and usable systems that solve actual problems.
My mindset is always problem-first and system-driven:
For me, model accuracy alone is never enough — I care about reliability, scalability, and real-world impact.
Theory explains what is possible.
Implementation teaches what is real.
Over time, my curiosity evolved into hands-on work across AI, ML, deep learning, and full-stack engineering.
I actively build:
I enjoy connecting AI models with real users, real APIs, and real constraints — not just training models in isolation.
That’s why I build my own:
Currently, many of these AI systems even run locally on my own machine — because understanding how things work at every level matters to me 😄
Strong engineers are not defined by tools, but by their ability to:
Tools change.
Frameworks change.
Problem-solving mindset doesn’t.
A lot of my personal philosophy can be summed up like this:
while problem_exists:
understand_problem()
design_system()
build_solution()
break_it()
fix_it()
improve_it()
🌱 Always Learning. Always Building.
I’m continuously:
This portfolio is not just a showcase — it’s my personal space on the internet 🌍
A place where I document my journey, explain how I think, and show how ideas evolve into working systems that people can actually use.
From playing with motors and fans as a kid, to building AI-powered platforms and intelligent systems today — the journey has been long, messy, and deeply rewarding. Every challenge, failure, and late-night debugging session has shaped how I approach problems now.
If my way of thinking resonates with you — if you believe in skills over shortcuts, systems over hacks, and building things that matter — feel free to connect 🤝
I’m always open to learning from others, collaborating on meaningful ideas, and working together to use technology to make things simpler, smarter, and better.
Because at the end of the day, technology is not about showing intelligence —
it’s about creating impact.
And honestly…
I’m just getting started 🚀🔥
“The best way to predict the future is to build it.” — Alan Kay
02/ Skills
Supervised & unsupervised learning, feature engineering, model evaluation, pipeline design.
CNNs, RNNs, Transformers, GANs. Tokenization, embeddings, Hugging Face ecosystem.
RAG pipelines, fine-tuning (LoRA/PEFT), tool-calling agents, inference optimization.
EDA, statistical analysis, data cleaning, visualization, feature selection.
Full-stack web with React/Next.js, Node/Express APIs, and relational/document stores.
Cloud, containers, messaging, observability, IoT, and model serving in production.
03/ Currently
No description provided.
~/work — system map
Selected Work
An MCP server that gives Claude full natural-language control over any REST CMS — 32 auto-generated tools, human approval gate, policy engine, semantic search, circuit breaker, and 78 tests. Works with Supabase, Strapi, Directus, Payload,…

"A full-stack agentic pipeline that scrapes 5 job boards daily, scores listings with a structured AI rubric, rewrites your resume per role, and submits — with two mandatory human gates hardwired in. Zero unreviewed submissions.

Scene Sorter is an AI-powered application that automatically classifies images into real-world scene categories such as mountain, forest, glacier, sea, street, and buildings. It supports both single and batch image uploads, organizes…

Selected Works
A measurement-first look at how multi-agent LLM systems should share context. Across 33 real Claude generations, reference-passing (TOAP) and a competent summarizing orchestrator both cut tokens with …
ContextForge: A Research-Grade Nexus for Secure, Persistent AI Memory Developed a "Defense-in-Depth" architecture to solve the "Stateless RAG Gap" in AI Agents. By implementing a dual-signal security …
PDFBuilt multiple end-to-end AI/ML projects focusing on real-world problem solving, including computer vision, recommendation systems, and predictive analytics. Worked across data preprocessing, model development, API design, and frontend integration.