Apr 13, 2026 · abhishekuniyal.com
Why this site is a publication, not a portfolio
A short note on why the site is structured like a publication with a volume number and a 'currently' stamp, instead of like a one-page portfolio.
Abhishek Uniyal
Vol. 01 · Apr 26 · Updated today
Currently — agent platform SDK · text-to-SQL engine
§ 01 Selected Work
Founding engineer · 2024 – present · SynergyBoat Solutions
A US EV startup scaled to 150+ B2B clients with ~40% lower dev cost and extended runway on a $2M seed.
TypeScriptMCPisolated-vmDenoRedisPostgreSQLBullMQ
Founding engineer · 2024 – present · SynergyBoat
Cut net-new boilerplate across SynergyBoat's AI products by extracting AST editing, token-budgeted context, and query-intelligence into one shared toolkit every internal agent now depends on.
TypeScriptts-morphpnpmTurborepo
Founding engineer · 2024 – 2025 · SynergyBoat client
Non-technical operators answered business questions without writing SQL, live across three databases.
TypeScriptPostgreSQLMySQLMongoDBOpenAI
Founding engineer · 2025 · SynergyBoat (internal)
Gate for AI-generated code that runs 95% of validations with zero LLM tokens, catching boundary violations agents introduce.
TypeScriptts-morphAST analysis
§ 02 Writing
Apr 13, 2026 · abhishekuniyal.com
A short note on why the site is structured like a publication with a volume number and a 'currently' stamp, instead of like a one-page portfolio.
Mar 10, 2026 · Medium
95% of the validations an AI coding agent needs can be done without calling an LLM. Here's the AST-based gate we built, and why it saved us both tokens and correctness.
Nov 22, 2025 · Medium
Intent classification sounds like an LLM problem. In practice it is a budgeting problem — what fits in 8K tokens, and what gets pruned before the model ever sees it.
§ 03 Notes
§ 04 How I Work
AI is a value-add, not a reason. The product still has to make sense to the customer without it.
I build AI-native systems because I want founders to ship products that actually work. Seven years across YC-backed startups and founding-stage teams have taught me that the difference between a demo and a system is whether it survives production load, real users, and constraints the pitch deck never mentioned. The work is technical; the reason it exists isn’t.
Fractional CTO engagements with founders shipping AI-native systems.
Currently open to selective fractional-CTO conversations.
§ 05 Contact
If something here made you think "I want to work with this person," the fastest way is an email.