Two senior founders · open for project work

We build AI systems
that survive in production.

Buildgent is a two-founder AI studio. We design, build and ship agentic AI, RAG and document-intelligence systems that hold up in production — by engineers who've cut client workloads ~80% with AI agents and scaled platforms to 9M+ users.

Built at — our prior roles ReBillion· Labcorp· AAVAA· TEKsystems· Leher Track record from our prior roles — Buildgent is our new studio.
0
work automated by AI
0
users scaled
0
peak DAU
0
faster AI APIs
0
continents
01 The studio

We're two engineers who've shipped AI to millions of users. One builds the brain, one builds the whole product. We ship systems that run in production and cut real cost — not slideware.

02 Capabilities

What we do together

Two complementary skill sets, one capability surface — from raw idea to model to product to scale.

Anshul lead

Agentic AI & LLM Systems

Planner–executor agents that plan, call tools (via MCP) and act — with retrieval, memory, guardrails and human-in-the-loop so they hold up on real work, not just demos.

LangGraphMCPTool-callingPlanner–executorHuman-in-the-loop
Anshul lead

Document Intelligence / OCR

Turn messy PDFs, scans and forms into clean, validated, structured data your systems can trust.

TextractDocument AIMistralLlamaParseIDP
Karan lead

Full-Stack AI Product Engineering

Complete AI products — LLM features (LangChain/LangGraph agents, RAG) on a React + Python/Go/Java stack, with ML models (TF/PyTorch) wired into the loop.

ReactLangChainRAGFastAPIPyTorch
Both

Backend & Distributed Scale

Microservices, multi-cloud, queues and observability — infrastructure proven to 9M+ users and 250K+ daily actives.

MicroservicesAWSGCPAzureKubernetesgRPC
Both

Fractional AI Engineering Lead

Senior architecture, technical direction, de-risking and team setup for companies adding AI to their product — without the cost of two full-time hires. We help you pick the right approach, prove it, and get your team shipping.

AI architectureModel selectionEvals & guardrailsDe-riskingTeam setup
03 How we build

AI systems, built to survive production

Not a chatbot demo — engineered agentic systems with retrieval, evals, guardrails and a human in the loop where it counts.

01

Planner–executor agents

Agents that decompose a task, call tools through a typed registry (MCP), and recover from failure — orchestrated with LangGraph.

02

Retrieval & memory

Grounded in your data: embeddings + hybrid search over a vector store, reranking and citations — so answers are traceable, not hallucinated.

03

Evals & guardrails

Golden datasets, LLM-as-judge and rubric grading gate every change in CI; structured outputs, validation and prompt-injection defenses on the edges.

04

Human-in-the-loop

On consequential actions the system pauses for review, with full traces and audit trails — autonomy where it's safe, oversight where it matters.

Eval-driven delivery
Nothing ships below the bar. We treat AI quality like tests — golden datasets, LLM-as-judge + rubric grading, and regression gates in CI — so behaviour is measured, not vibes.
Trust & safety
Your data stays yours — we don't train on it.
PII redaction & field-level validation.
Human-in-the-loop on consequential actions.
Guardrails, citations & audit trails.
References from past teams available on request.
04 Founders

One builds the brain. One builds the whole product.

AG

Anshul Goel

Agentic AI · Document Intelligence · Backend Scale
Engineering Lead @ ReBillion
Bangalore, India

Builds the brain — autonomous AI systems that read messy real-world documents and act on them. Grew from intern to tech lead scaling platforms to millions before going deep on agentic AI.

  • ~80% client man-hours cut with agentic AI
  • Agentic control plane — LangGraph · Vertex AI · OCR/IDP
  • Scaled platforms to 9M+ users · 250K+ DAU
View portfolio ↗
KS

Karan Singla

Senior AI Full Stack Developer · LangChain/LangGraph agents · RAG · ML (TF/PyTorch)
Senior Dev @ Labcorp · M.Eng, Concordia University, Montreal, Canada
Bangalore, India

Builds the whole product — LLM agents and RAG on top, React + Python/Go/Java underneath, ML models in the loop. 8+ years shipping AI products across two continents, with a Master's in Software Engineering.

  • GenAI in production — LangChain/LangGraph agents · RAG
  • 65% faster API responses · 50× data scale
  • Mentored 10+ engineers
View portfolio ↗
05 The journey

Two paths. One studio.

We started at the same university and the same first company — then the road forked across two continents, and brought us back as co-founders.

2015 2026 ANSHUL KARAN
2015
Chitkara University
2018–19
Leher · early careers, together
2019–22
Scaled to 9M+ users · 250K+ DAU
2023→
Eng Lead, agentic AI @ ReBillion
2022–24
M.Eng, Concordia · Montreal
2023→
AI Full-Stack — AAVAA, now Labcorp
2026
Reunited — Buildgent
2015 · 2018–19

Shared start

Both at Chitkara University, then both at Leher — early careers side by side.

Anshul · India
2019–22

Scaled to 9M+

250K+ DAU, 100+ microservices.

2023→

ReBillion

Eng Lead, agentic AI.

Karan · Montreal
2022–24

M.Eng, Concordia

Moved to Canada.

2023→

AAVAA → Labcorp

Senior AI full-stack.

2026

Reunited — Buildgent

Two paths converge into one studio.

Built together from the start. Same university, same first company — we've been shipping side by side since before this studio existed.

06 Engagements

Proof, not promises

The systems we've led and shipped — the studio's track record, told in challenge, approach and result.

ReBillion · Agentic AI · Anshul-led

Autonomous AI for U.S. Real Estate

Agentic AI control plane · 2023–present
Challenge
Real-estate transactions buried in manual coordination — signatures, deadlines, legal PDFs nobody wants to read.
Approach
An agentic control plane (LangGraph + Vertex AI, tool-calling via MCP) over an OCR/IDP stack (Textract, Document AI, Mistral, LlamaParse) — with field-level validation, citations and human-in-the-loop on consequential actions.
Result
~80% fewer client man-hours — clients close more deals without scaling headcount.
80%
manual coordination removed
Multi-agent
LangGraph · Vertex AI
4 engines
OCR / IDP stack
Leher · Scale & Systems · Anshul-led

Scaling a Social Platform to Millions

Multi-cloud infrastructure · 2018–2023
Challenge
Build and scale audio-video social products for a fast-growing, massive audience — reliably and cheaply.
Approach
100+ microservices across AWS, GCP & Azure, a 10M+/day push system, on near-zero cloud cost via startup credits.
Result
9M+ users, 250K+ peak DAU, a $100K/month service and ~$1.2M ARR — on near-zero infra (startup credits).
9M+
users served
250K+
peak DAU
~$0
infra (credits)
AAVAA · AI Full-Stack · Karan-led

AI-Powered Full-Stack Platform

React + Python/Go · Montreal · 2023–2025
Challenge
Ship full-stack apps with real-time AI decisioning and conversational experiences — fast and production-grade.
Approach
High-performance REST & gRPC services, RAG and LLM agents (LangChain/LangGraph) plus ML (TF/PyTorch), on AWS + Kubernetes with CI/CD.
Result
65% faster API responses with real-time AI features running in production.
65%
faster API responses
Full-stack
React · Python · Go
RAG
agents in prod
07 Process

How we work

A clear, low-risk path — you see value before you commit to the whole build.

01

Frame

We find the real problem and the workflow behind it — not just the feature request.

02

Prototype

A working POC, fast — so we validate before anyone invests in the full build.

03

Harden

Eval-driven: golden datasets, LLM-as-judge, regression gates in CI, guardrails and edge-case coverage — the part most demos quietly skip.

04

Hand off

Docs and knowledge transfer, so your team fully owns what we built.

Typical engagement: focused 6–8 week sprints — boutique cadence, senior hands on every line.

08 Toolkit

One combined stack

Both founders' tech, merged into a single studio toolkit across the whole pipeline.

Models

ClaudeGPT-4-classGeminiLlama / QwenMistralModel routing

Agentic & LLM

LangGraphLangChainMCPTool-callingMulti-agentGuardrails

Retrieval / RAG

EmbeddingspgvectorPineconeHybrid searchReranking

Evals & LLMOps

LLM-as-judgeGolden datasetsLangSmith / LangfusePrompt versioningToken-cost

Document Intelligence

TextractDocument AILlamaParseIDP

ML

TensorFlowPyTorchScikit-learn

Frontend

ReactAngularTypeScript

Backend & Languages

PythonGoJavaNode.jsgRPC

Cloud & DevOps

AWSGCPAzureDockerKubernetesCI/CD
09 Why us

Why this studio

01

Two senior owners

No junior hand-offs, no account managers. You work directly with the people who build it.

02

End-to-end coverage

Idea → model → product → scale. Two complementary stacks that cover the whole pipeline, no gaps.

03

Production, not demos

Systems that run in production and cut real cost — proven at millions of users, not in a notebook.

04

Continuity & focus

Two people who can both own the work, taking on a deliberately limited number of projects at once.

What should we build together?

Tell us what you're shipping. The first 30-minute call is on us.