Saras
Updated 13 Aug 2026
AI Engineer (Contract)
Saras is a learning system for busy professionals — an AI tutor that lives in your pocket via Telegram, starting with GMAT prep and expanding to GRE and other standardized tests, with WhatsApp and iMessage on the roadmap. It's built on Hermes Agent, with a profile-based architecture: every new user gets their own git-hosted Hermes profile, provisioned through an automated pipeline — build once, clone infinitely.
We shipped v0 in late July and are live in early access: 15+ users on trial today, with a few more waitlisted, and on track to hit our first $1000 in revenue by the end of August. We're targeting a $120B+ test prep market, starting with US/EU GMAT aspirants. Priced at $150-250 for 3-month access, with a 14-day free trial. Now that the core loop works end to end, the focus is making the platform scale cleanly past a handful of hand-held users, hardening the provisioning pipeline, automating onboarding fully, and wiring up payments.
What we build through Saras feeds forward into Sanganak Works' larger ambitions. We're independent, bootstrapped, remote-first, and mission-oriented. Very early, with room to experiment and bring frontier tech to those who deserve it. Read about our manifesto here.
What you'll actually do
This is not a traditional engineering role. This is 80% configuration, 15% coding with AI agents, and 5% tasteful intuition about systems. You'll work alongside AI agents daily — Hermes, Codex, and the rest of the stack are your co-workers. This is how engineering will get done in the future. Be ready for it.
Channel expansion: WhatsApp & iMessage
Saras lives on Telegram today. You make the "text one number, get a personalized tutor" experience work end-to-end as we open it up to WhatsApp and iMessage, building the router/bridge that holds the stream and dispatches to the right profile with hard memory isolation per user, regardless of channel.
Payment gateway
Payments are already underway; the frontier is agentic payment rails: in-line purchases, price negotiation, and transacting on the user's behalf inside the conversation. This isn't a one-time gateway to wire up and forget; it's making payments a native part of how Saras coaches.
User onboarding flow
From "user texts the number" → profile provisioning → conversational intake (target score, weak areas, study timeline) → personalized study plan. Must happen in one message round-trip, zero manual ops.
Integration engineering: 3P data sources & Google Calendar
Wire Saras into third-party data sources and Google Calendar so it reads the user's real schedule, pulls in relevant context, and plans study around an actual life, not a generic template.
Proactive check-in/check-out system
Cron-triggered nudges per user profile. Saras doesn't wait for the user to initiate; it checks in daily, tracks progress, adjusts the plan. This is what makes it a coach, not a chatbot.
Monitoring & ops
Per-profile logging, memory integrity checks, gateway health dashboard. One VM, one Docker project, one gateway process, and you keep it running.
Open source contributions
We build on the shoulders of giants, backed by plenty of our own elbow grease, making scalable systems that support everyone. Our stack is largely open-source projects.
What you'll work with
The full architecture is a conversation for the interview; for now, the keywords that matter: autonomous AI agents, multi-tenant messaging, and cloud deployment. Think Docker, cron-driven automation, and agent frameworks in the Hermes / OpenClaw lineage. We're open floor on the rest; if you can reason about self-running agents and production cloud systems, you can pick up our specifics on the job.
What this role is not
- Not a research role: we ship production systems, not papers.
- Not a pure backend role: you'll configure agents, write docs, make product calls.
- Not a "senior architect" role: you'll own decisions but also write the code.
Fluency over credentials
- 2+ years working on production-grade systems
- 1+ years building or contributing to AI projects (agent frameworks, LLM integrations, or autonomous systems)
- Fluency with agent memory, self-continuing loops, agent skills (how agents actually work, not just API calls)
- Comfortable writing specs and documenting decisions
- Exposure to Hermes Agent or OpenClaw framework
- Active GitHub profile with real projects (matters more than a resume)
- Experience with multi-tenant or multi-profile agent systems
Compelling GitHub profiles or portfolios will be preferred over swanky resumes.
Compensation & terms
- Duration
- 3 months initial, option to extend to 6 by mutual agreement
- Cash
- ₹35,000/month
- Profit share
- 15% of Saras's net profit (paid quarterly, kicks in after ₹1L+ MRR)
- Tooling
- All AI tools and subscriptions fully expensed (Claude Max, Codex, any API credits you need)
- Work style
- Remote or Noida-based. Low-BS environment — just you, the founder and a couple of other high-agency builders (ex-Atlassian, Polygon, Salesforce), and AI agents. Be on your toes, learn and share your findings. Move fast, no bureaucracy.
After the 3-month period, we're open to renegotiate terms based on product trajectory and performance.
- Codex/OpenRouter/Claude subscriptions, based on what you need
- Blog/X visibility: you write about what you build on Masala Dew's Substack and X (or build your own!). A real career asset.
- Shipping experience: you architect systems from zero that handle real users, real scale
- Direct influence: every architectural decision from day one
- Frontier exposure: building at the edge of what AI agents can do, not maintaining CRUD apps
Apply on BITSians. Include your GitHub profile or portfolio. We'll read it before we read your resume.