Dan Abramov tweet
· Twitter / XReact's creator (and now Bluesky engineer) tweeted about a production React app I shipped at Codebrahma, back when shipping non-trivial React in production was still rare.
Open to founding-engineer roles
I ship any product
Product Engineer · AI · 11 years
I ship any product
with AI in every layer,
from infra to DAGs to APIs to UI.
First engineer at PortfolioPilot. Founding engineer at IntuitionAI (acquired by Domino Data Lab); shipped early at Navya, Rippling, AthenaHealth. Built one of the first ChatGPT plugins. Today: shipping AI agents in production. And using agents to ship faster than my old self ever could.
- 0
- years shipping
- 0
- companies · YC · Sequoia · big VCs
- 0
- startup acquired
- ∞
- AI features in prod
/ shipped at
AthenaHealth Codebrahma Rippling Navya IntuitionAI Domino Data Lab Global Predictions PortfolioPilot AthenaHealth Codebrahma Rippling Navya IntuitionAI Domino Data Lab Global Predictions PortfolioPilot
/ career.dag
Eleven years, rendered as a directed acyclic graph.
It's how I think about systems. And, it turns out, how my career has run too. Click any node to read the case study.
dag_id=mohith_g · schedule=daily · running
v11.12
running · today flagship chapter
/ selected work
Three case studies. No filler.
Each one walks through the setup, what I owned, the hard problem I solved, and the outcome.
/work/portfoliopilot
First
engineer · quant engine + AI surface, ground-up
2021–now·Founding / First Engineer
Global Predictions
SEC-registered AI financial advisor. Hedge-fund-grade quant engine, built ground-up. The AI is the legibility layer.
First engineer on the team. Built the SEC-registered AI financial advisor ground-up. Hedge-fund-grade quantitative engine underneath; AI on top is the layer that turns dense statistical output into language a user can act on.
Quant infraSEC-registeredAILLMsPython
Read the case study
/work/intuition-ai
→
acquired by Domino Data Lab
2017–2018·Founding Engineer
IntuitionAI
Founding-engineer story: model monitoring, before it was a category.
Built model monitoring before "model monitoring" was a category. Complete system on Kafka, Python and React. Drift, data quality, eval. Acquired by Domino Data Lab in 2018.
ML monitoringFoundingKafkaPythonReact
Read the case study
/work/rippling
2016
production React when production React was rare
2016–2017·React Engineer (via Codebrahma)
Rippling
YC HRIS, early days. The React work that earned a Dan Abramov tweet.
Built early product surfaces for Rippling, Parker Conrad's next company after Zenefits. Worked through Codebrahma, the React shop where I shipped my first React app to production.
ReactYCHRISFrontend craft
Read the case study
/ the other chapters
The supporting cast. Each one shaped how I work.
Domino Data Lab
2018–2021Senior Software Engineer
Joined post-acquisition. Worked on Domino's core platform, built real-time systems, and led an architecture revamp of significant parts of the codebase.
Distributed systems MLOps Real-time Scala React K8s
Codebrahma
2015–2017React Engineer
React dev shop. Shipped my first production React apps here (Lendwell, then Rippling). Got a tweet from Dan Abramov for the work.
React Open source
AthenaHealth
2015–2015Software Engineer
First job out of college. Healthcare software at scale, written in Perl. Learned what real engineering rigor looks like.
Healthcare Perl
/ what i build
Six surfaces. All of them shipped, in production.
I'm a Product Engineer. I don't get squeamish about which layer of the stack I'm in. The question is always what does the product need next?
/ai
AI features in production
Real LLM features that sit on real systems, not LLM wrappers around someone else's engine. Tool-using agents, streaming UIs, end-to-end coherence between substance and surface.
- PortfolioPilot AI translation layer over a hedge-fund-grade quant engine
- Early ChatGPT plugins → GPTs migration → custom agent runtime
- Streaming UIs over LLM tool calls, citing the source the engine ran on
/realtime
Real-time systems
Sub-second feedback loops where they matter, from real-time model monitoring at Domino to live-streaming AI advice at PortfolioPilot.
- Model drift / data-quality monitoring at scale
- Streaming AI responses with optimistic UI
- WebSocket + SSE infra and the gnarly state-sync that comes with it
/pipelines
Data pipelines & DAGs
Daily ETL, financial-data ingestion, multi-vendor normalization. The unglamorous plumbing that the entire quant engine sits on.
- Multi-source financial data ingestion at PortfolioPilot, normalized into one canonical model
- Macro-signal pipelines feeding multi-model orchestration
- Celery, Airflow-style DAGs, idempotency, replay safety
/frontend
Frontend craft
React since the early days. First React-to-prod app earned a Dan Abramov tweet. Today: design systems, complex state, taste.
- PortfolioPilot product UI: design system + component library
- Earned a Dan Abramov tweet for production React work in 2016
- Server components, Suspense, modern patterns shipped to real users
/backend
Backend & data
Python + Node services, Postgres, schema migrations, the full lifecycle. Nothing gets thrown over a wall.
- Python services (FastAPI, Django) and Node services (Express, Hapi)
- Postgres schema design, partitioning, query plans
- Authn/authz, multi-tenancy, financial-data correctness
/infra
Infra & scale
AWS, Kubernetes, observability, cost. I can stand it up, scale it, and pay for it.
- K8s deployments, GitOps, blue/green
- Observability (traces, metrics, structured logs, end-to-end annotated outputs)
- Cost-aware architecture; choosing managed vs. self-hosted with eyes open
/ moments
A few landmarks along the way.
-
2016 -
2018 IntuitionAI → Domino Data Lab acquisition
· Domino Data LabThe model-monitoring company I co-built as founding engineer was acquired by Domino Data Lab. Joined Domino's core platform team for the next three years.
source -
2023 Early ChatGPT plugins for PortfolioPilot
· PortfolioPilotBuilt one of the first wave of ChatGPT plugins: financial-advice retrieval over PortfolioPilot data. Migrated to GPTs as the platform evolved, then to a full agent stack.
source
/ talk to me
Building something that needs a founding engineer?
I'm open to founding-engineer or first-AI-engineer roles at AI-forward seed/Series A teams. Also happy to talk shop about agent stacks, building real systems under the LLM, or how to ship AI features that don't embarrass you.