AI & software consulting · Thiruvananthapuram, India
AI systems that earn
their place in production.
We build AI and automation software that runs live: trading systems, LLM services, data platforms. Then we measure whether it actually works.
- 01Multi-broker trading automation
- 02Local & hosted LLM services
- 03Evaluation-first delivery
- 04Verified against reality
Tools we work with
- Claude
- OpenAI
- Ollama
- Llama
- Mistral
- Python
- FastAPI
- TypeScript
- Node.js
- Next.js
- React
- MongoDB
- PostgreSQL
- Prisma
- Interactive Brokers
- Tradovate
- Alpaca
- TradingView
- Pine Script
- Docker
- GitHub Actions
- AWS
- Vercel
What we do
Six things we do well.
We stay narrow on purpose. Each of these is something we have built, shipped and operated, not a service line added to fill a page.
Applied AI & LLM engineering
Retrieval, structured outputs, agents, fine-tuning and local deployment, with an evaluation harness so you know what improved.
Trading & fintech systems
Automated execution, broker integrations, reconciliation and risk controls for futures, options and equities.
- Broker integrations: Interactive Brokers, Tradovate, Alpaca
- Futures, options (including SPX 0DTE) and equities
- Verified execution and broker-to-ledger reconciliation
- Risk guardrails: daily caps, give-back protection, kill switches
- Strategy research, backtesting and shadow-mode validation
Data platforms & pipelines
Ingestion, time-series storage, replay and analytics that keep decision systems honest.
Product engineering
Full-stack web applications and cloud deployment, built to be maintained.
Speech & Indic-language AI
Transcription and language pipelines for Malayalam and other Indian languages.
AI adoption & automation
Putting Claude and GPT-class models into everyday workflows, with measurement built in.
Selected work
Systems we have built and run.
Some of this is client work, some is our own production infrastructure. All of it is live.
Multi-broker automated trading platform
An unattended trading system running index futures, SPX 0DTE options and equity strategies across three brokers through the full US session.
- Brokers
- 3
- Asset classes
- Futures · Options · Equities
- Session
- Unattended
Local LLM market-analysis service
Replacing a paid model API with an open-weights model on a laptop-class machine, without giving up rigour.
- Inference
- On-device
- API cost
- → 0
- Evaluation
- Replayable
Multi-entity wealth tracker
A family-office application tracking assets across the United States and India for several users, in one consistent view.
- Users
- Multi
- Markets
- US · India
- Ops
- Zero
How we work
Weeks to a working slice. Evidence before scale.
The shape of every engagement is the same: agree what success means in numbers, build the thinnest thing that can be measured, then harden what earns it.
- 011–2 weeks
Discover
We map the workflow, the data and the decision being made. We agree, in numbers, what a successful outcome looks like before writing code.
- 022–6 weeks
Prototype
A thin, end-to-end slice on your real data, alongside a simple baseline. If the AI does not beat the baseline, you find out now, cheaply.
- 03Ongoing
Harden
Evaluation suites, guardrails, reconciliation, observability and cost controls. This is the work that turns a demo into something you can rely on.
- 04Your call
Operate
We run it with you, review it weekly, and either hand it over to your team or keep operating it. No lock-in either way.
Principles
Evidence over hype
Every model change is graded against a baseline. A result that says 'the model adds nothing' is a good result.
Production is the goal
Notebooks are where ideas start, not where they end. We build for the 3 a.m. failure, not the demo.
Senior engineers, small teams
You work directly with the people writing the code. No hand-offs, no account layers.
Open-weights when it makes sense
Hosted frontier models where quality demands it, local models where cost, privacy or latency wins.
Cost is a feature
Caching, dedup, routing and right-sizing are part of the design, not an afterthought.
Verify, don't assume
In trading we confirm every action against the broker. That same discipline shapes everything else we build.
Cedarwood AI Engineering Internship · Thiruvananthapuram
Learn by shipping something real.
A project-based internship where small teams build a complete, deployed application from an empty repository, using the same tools and review standards we use for client work.
- Duration
- 4–8 weeks, project-based
- Team
- 2–3 interns per project
- Workflow
- Jira, GitHub, reviewed pull requests, CI
- Outcome
- Deployed app, demo, retrospective
AI-automated swing trading for NSE stocks
Three engineering interns built a paper-trading swing system for Indian equities from scratch: daily market-data ingestion, a technical screener, a paper-trade lifecycle with stop and target management, and a live dashboard, deployed to the cloud.
Contact
Tell us what you are trying to build.
A short email is enough. We reply within two working days with either a clear next step or an honest reason we are not the right fit.
Based in
Cedarwood Consulting Pvt Ltd
Thiruvananthapuram, Kerala, India