Website:
transfrm.in
Job details:
Transfrm Labsby Rachitt Shah [Index][Services][Blog][evalOS][Contact]
IndexServicesBlogevalOSContact
rachitt@transfrm.in
// APPLIED AI RESEARCH + CONSULTING LAB
Applied AI Systems& Strategy
Transfrm Labs is an applied AI research and consulting lab.
Most AI systems work in demos. They break in production.
I help teams close that gap: proper architecture, real evaluation systems, and infrastructure that survives actual traffic.
Not theory. Not decks. Systems that ship.
Explore evalOS
// Currently booking Q2 2026. Limited availability.
SCROLL →
// SIGNAL_LOG: Trusted by teams who ship:
// From Fortune 500 to Series A. The common thread: they needed it to work, not just demo.
// WHAT_I_DO: Three things. Done well.
/ 01
Production Architecture
Most teams ship the demo and call it done. Then production happens.
I Build Systems That Survive Real Traffic:
- →RAG pipelines that don't hallucinate at scale
- →Vector databases handling actual QPS, not benchmarks
- →Evaluation loops that catch failures before users do
- →Infrastructure your on-call won't curse at 3am
Why this matters: The gap between "works locally" and "works at scale" is where most AI projects die. I've shipped LLM systems since GPT-3. I know what breaks.
// Pro tip: If you can't measure it, you can't improve it. Evals first, features second.
/ 02
Technical Due Diligence
Most AI moats are demos with good lighting. I find the real ones.
For investors and acquirers who want to know what's actually under the hood:
- →Architecture reviews beyond "they're using GPT-4"
- →Team assessment: can they ship, or just prototype?
- →Technical debt mapping before it becomes your debt
- →The questions founders hope you won't ask
Why this matters: Bad diligence means you inherit someone else's mistakes. I've done this for funds deploying real capital. I know what to look for.
// Caveat: I'll tell you what I find, not what you want to hear.
/ 03
Fractional AI Leadership
You need senior technical direction. You don't need another $500K salary.
I Embed With Your Team During High-growth Phases:
- →Architecture decisions that won't need unwinding in 18 months
- →Hiring support: who to bring in, who to pass on
- →The hard conversations your team won't have without air cover
- →Strategic direction when a full-time AI lead isn't yet required
Why this matters: The gap between product vision and engineering reality kills companies. I've sat in the seat. I know what breaks when leadership is absent.
// Try this: Before hiring a full-time AI lead, ask if you actually need one yet.
// FROM THE LAB: research, hardened into product.
Private Beta — 2026
evalOS
Evaluation-as-a-service for production AI.
One line of code turns on OpenTelemetry-native observability, automated JudgeLM evaluation, and human-in-the-loop quality gates. The eval layer I kept rebuilding for clients, now a product.
OpenTelemetry-nativeJudgeLM evaluationDashboards + costEvalGround
Explore evalOS
evalos — production.eval
// ENGAGEMENT_PROTOCOLS
PROTOCOL: SECONDMENT
Embed & Execute
Long-term problems need long-term presence. I integrate with your team: standups, PRs, Slack, the lot. Not a consultant who hands off a doc and disappears. I stay until it ships.
What You Get:
- → Direct integration with your engineering team
- → Knowledge transfer built into the engagement
- → Architecture decisions made with context, not assumptions
Minimum commitment: 3 months.
Best for: Core AI product builds, architecture overhauls, teams in transition.
PROTOCOL: CONTRACTING
Target & Resolve
Not every problem needs an embedded engagement. You have a specific issue. I scope it, solve it, and hand it back. No ambiguity, no scope creep. You get the deliverable, not a deck about the deliverable.
What You Get:
- → Fixed scope, fixed timeline, fixed price
- → Direct access: no layers, no account managers
- → Clean handoff with documentation
Minimum commitment: 2 weeks.
Best for: Evaluation systems, pre-fundraise technical cleanup, unblocking specific bottlenecks.
// Non-standard arrangements available for the right problems.
// EXTENDED_ROSTER: When the mission needs more firepower:
Rajaswa Patil
Applied AI. Ex-Postman, Microsoft Research.
Anshul Bhide
AI Strategy. Ex-Replit, VC. MIT Sloan.
Ayush Chaurasia
ML/CV. Co-founded Ultralytics (YOLOv8).
Kumar Shivendu
ML Platforms. Qdrant, Search, Devtools.
Vipul Maheshwari
Applied ML. LanceDB, Superlinked.
Rohan Balkondekar
GenAI Lead, Air Arabia. F500 + YC.
// Not contractors. Collaborators. Each one could run their own shop.
// READY_TO_DEPLOY?
I take on limited engagements. If it's not a fit, I'll tell you. If it is, we move fast.
// Or skip the form: rachitt@transfrm.in
// Response time: 48 hours. No sales process. Direct conversation.
Transfrm Labs
by Rachitt Shah
// Applied AI systems. Production-grade.
// Building with teams at Accel, Sequoia, and friends.
// Bangalore
Connect
- [ Email ]
- [ LinkedIn ]
- [ Twitter ]
- [ GitHub ]
Legal
© 2026 Transfrm Labs. All rights reserved.
// SYSTEM_STATUS: OPERATIONAL
Click on Apply to know more.