MARK XIV · BUILD IN PROGRESS
STILL IN THE CAVE
Final-year ECE at NIT Agartala. I build AI agents, trading systems, and research pipelines. Fourteen builds in three years. Most of them taught me something. Some of them worked.
02 / HALL OF ARMOR
Fourteen suits in three years.
Every project is a Mark. The failures are on the wall next to the wins, because leaving them off would make the wins mean less.
03 / SYSTEMS DIAGNOSTIC
Subsystems, honestly rated.
A row of 95%-filled bars is the least credible thing on a portfolio. The low readings below are the reason you can trust the high ones.
POWER SYSTEMS
AI AGENTS & LLM ORCHESTRATION
- Google ADK / multi-agentPRIMARYGoogle ADK / multi-agent: PRIMARY, 80 percent
- RAG + retrievalOPERATIONALRAG + retrieval: OPERATIONAL, 67 percent
- LLM fine-tuning (QLoRA)OPERATIONALLLM fine-tuning (QLoRA): OPERATIONAL, 53 percent
- MCP / agent infraPRIMARYMCP / agent infra: PRIMARY, 73 percent
- Gradient boosting ensemblesPRIMARYGradient boosting ensembles: PRIMARY, 73 percent
TARGETING
FINANCE & ECONOMICS
- Fundamental / equity researchOPERATIONALFundamental / equity research: OPERATIONAL, 67 percent
- Personal financePRIMARYPersonal finance: PRIMARY, 73 percent
- Venture & startup economicsOPERATIONALVenture & startup economics: OPERATIONAL, 60 percent
- Macro & market structureDEVELOPINGMacro & market structure: DEVELOPING, 47 percent
AIRFRAME
ENGINEERING
- PythonPRIMARYPython: PRIMARY, 80 percent
- React / Next / TailwindOPERATIONALReact / Next / Tailwind: OPERATIONAL, 60 percent
- FastAPI / PostgresOPERATIONALFastAPI / Postgres: OPERATIONAL, 60 percent
- DSADEVELOPINGDSA: DEVELOPING, 47 percent
04 / CAVE · POWER UNSTABLE
The first suit gets built in a cave. With a box of scraps.
Everybody remembers the flying. Nobody remembers that the first one couldn’t land.
So here’s the honest version.
I’m running four things at once right now. A startup with a deadline I set myself. Campus placements. Building in public. Content. I don’t know which one works. I’m not going to pretend I do.
What I do know is this: in three years I’ve shipped fourteen builds — flower delivery, hospital compliance, clinical triage, equity research, autonomous research pipelines, Postgres tooling, agent infrastructure. Some placed in national competitions. Some got merged into other people’s repos. Some died quietly and deserved to.
People will tell you that’s unfocused. I’ve decided it isn’t. It’s a search — and I’m running it in public, and I’m running it fast, because the only way I find out which one is the real suit is to build all of them.
I’m confused. I know I’m confused. I’m doing it anyway.
That’s who I am.
Mark XV is next.
05 / CURRENT BUILD
What’s powered on right now.
MARK XIV · AGENTRONICS
ACTIVEWebMCP infrastructure — auth, observability, memory, and context management for AI agents interacting with websites.
- BUILD PHASE
- Infrastructure
- CADENCE
- Daily, alongside placements
MARK XV · IN FABRICATION
UNSTABLECurrently deep in AI agents, LLM systems, and interview prep. Open to full-time and internship roles in AI/ML engineering.
- GRADUATING
- May 2027
- SEEKING
- AI/ML engineering
COMMS · CHANNEL OPEN
LET’S BUILD
SOMETHING
Open to AI/ML engineering roles. Also just open to talking.
EMAIL
neelakandannithin@gmail.com ↗Long-form. Proposals, roles, decks.
X
@NeelakandanNC ↗Thinking out loud. Fastest replies.
LINKEDIN
neelakandan-nc ↗Formal outreach and recruiting.
GITHUB
NeelakandanNC ↗The source for most of the Marks.
YOUTUBE
@NeelakandanNC ↗Build logs, one episode at a time.