Data integration and agents, in one loop. Build it with us.
Conscience Technology fuses data integration and agent execution into a single loop, implementing Self-Improving Learning. We build systems that safely govern agents in real enterprise and public environments, and learn from human feedback.
We are reinventing the audit and maintenance risk that comes with probabilistic AI.
This system is already being adopted by a P&C insurer with 10B+ KRW in revenue, a commerce company at 50B KRW, and a bio company at 500B KRW. A US enterprise and one of the world's top study-abroad consulting firms are also preparing to adopt it.
Large enterprises and the public sector are conservative and cost-rational. They scrutinize whether AI agents can truly be governed, and weigh maintenance long before adoption. Our product is built exactly for that scrutiny.
These systems still need engineers who can deeply understand and integrate enterprise systems by hand. They need a front-view. We're hiring the kind of person who can help a company actually wield AI agents, keep them under control, and maintain them sanely.
Bootstrapped to 600M KRW in revenue and 20%+ operating margin in our first 6 months. This year we're going for 1.1B KRW, expanding aggressively.
Our Self-Improving system isn't chasing flash. It's looking at the essence of enterprise. Maintenance has to be overwhelmingly easy and rational, and adoption itself has to be low-risk. That's what we've learned in this industry.
The founders bring a mix of VC, AI research, and a non-CS engineer who went from zero to overwhelming growth in two years. We have a practical, excellent advisor network. The way we work can be relentless, but the speed at which we reach results is overwhelming. We've shortened a large-enterprise project by 70% of its original timeline.
For over a year after founding, we built 10+ products together in one apartment, failed at all of them, learned something new, and quickly sold a few. Come work with us and even if you fail you'll learn how to keep a steady heart, how to spend easy time with people you love, and how not to starve in a capitalist market. If you're a good person, you're welcome. A little short on skill is fine. We want to run together for a long time.
We have a long road ahead, and we want to bring great people in and figure it out together, learning every day. Our team genuinely dislikes office politics and behind-the-back talk. Sometimes a teammate surprises me: "You said that to their face and made up in 30 seconds?" That's the charm of this team. Transparent, honest, warm. We see ourselves as astronauts trusting each other in a vast, frightening empty space — fierce, but serious about being happy too.
From Korea we're expanding verticals fast — and now Japan, Singapore, and the US are next. We think we're a little different from typical data integration / SI / AX shops: deep technology, genuine care for customers, and a daily intensity that refuses to fall behind in this market.
Founding Solution Engineer (FSE)
KT Songpa Tower, Jamsil (partner office) or Gyeonggi Startup Innovation Center (5 min from Byeollae), and client sites
We build Nora — a solution for data integration, agent execution, hallucination detection, human feedback, and self-improvement. As we roll Nora out to enterprises, customers still hit many bottlenecks. Nora is fundamentally an agent server, so it has to be paired with thoughtful software-usage design: how to invoke it, where to run it, and how to actually get the most out of the agent. This question has to be asked seriously.
The essence isn't selling Nora. It's solving the customer's problem, with depth.
We see customers as companions for the journey. They trust us, and they're the channel through which we grow. That doesn't mean the customer is always right or that we should believe in the impossible. We just believe an optimistic attitude toward customers will make us happier in the end — and we want to travel with that belief.
What you'll do
- Analyze and define the customer's processes and problems
- Design the AI agent and data integration system
- Build and operate solutions on top of Nora
- Implement LLM, RAG, workflow, and agent systems
- Integrate the customer's internal systems and SaaS
- Identify and remove bottlenecks in live AI operations
- Turn customer feedback into product improvements
- Propose new features and product directions
- Run and manage the customer's AI transformation (AX)
Requirements
- Hands-on software engineering experience
- Has driven a problem end-to-end, from definition to solution
- Picks up new technology quickly
- Can translate a customer's problem into a technical one
- Communicates and collaborates clearly
- Strong execution and a deep sense of ownership
Nice to have
- Startup experience
- Experience on generative AI / LLM projects
- Python, TypeScript, SQL
- Built RAG, agent, or workflow systems
- Cloud operations on AWS / GCP / Azure
- Has carried a project while talking directly to the customer
- Consulting, solution engineering, or PM background
- Technical depth paired with business sense
Member of Technical Staff (MTS)
KT Songpa Tower, Jamsil (partner office) or Gyeonggi Startup Innovation Center (5 min from Byeollae)
Nora is the core of everything we ship. MTS is the role that builds that core directly. The data integration layer, agent runtime, hallucination detection, RLVR reward functions, sLM distillation — wherever you start, it all wires back into the same product.
The single thing we spend the most time on is designing RLVR reward functions well. Define a verifiable reward, refine it to the domain, train it directly on sub-4B sLMs, and verify the model actually learns from that signal. Where small models beat larger ones is decided here, and this is the one craft we want to do better than anyone. On top of that, you advance the causal reward layer with us — the deepest area we are pushing.
This is not a research-only role. You take a hypothesis to a model, validate it on evaluation sets, and ship it into a live customer environment. The cases FSE brings back from the field are absorbed here as product generalizations.
What you'll do
- Design and build Nora core components (agent runtime, data integration, hallucination detection)
- Design RLVR reward functions, build evaluation sets, train and distill sLMs
- Research and verify the causal reward layer
- Run the tracing-driven distillation pipeline for domain-specialized sLMs
- Absorb customer cases from FSE into product generalizations
- Automate evaluation sets and regression tests
- Extend internal tooling (Orbit, etc.) and review PRs
- Write internal tech notes — the seed of external publications
Requirements
- Software or ML engineer who has shipped and operated production systems
- Strong Python. Hands-on PyTorch and distributed training (FSDP, DeepSpeed, accelerate)
- Has turned a hypothesis into code and produced their own result
- Owns decisions in their domain
- Documents results so the next person can pick up cleanly
- Strong execution and a deep sense of ownership
Nice to have
- Strong in one or more of RL, RLHF, RLVR, distillation
- Strong opinions on hallucination detection, causal reasoning, or knowledge graphs
- Experience building agent runtime or orchestration systems
- Has reproduced and modified papers to produce your own result
- PhD or ABD (we look at what you've built more than your degree)
- Open-source contributions
We prefer to talk about visible results. We move in short, clear weekly cycles, and we work on top of our own ERP.
Hours
Weekdays: start between 8–10 AM, leave any time after 4 PM. You manage your hours within your contracted fixed-overtime band. We prefer conversations centered on visible output. (Customer schedules may shift this.)
Monday 10AM — Commitment Meeting
We declare in one line: what goal we'll hit this week, why it matters, and how we'll prove it.
Example · Commitment: Achieve high satisfaction grounded in real researcher use
Why: The customer counterpart reported significant pain
Delivery: User story #34, #35 completed — delivered as Gif
Wed mid-share · Fri 8PM Demo
Wednesday we share where we are against Monday's declaration. Friday 8 PM is Demo Meeting — present the working output.
Software we use
We run on Orbit, our in-house AI Native ERP — open-sourced internally. Need a feature? Open a PR; a lead reviews and merges. PMs, designers, and ops contribute directly. Communication is on Discord by default, with customer-specific channels as needed.
CV review
Response within 1 week
Google Meet chat (60 min)
A relaxed conversation with a Head or the CEO
Take-home (optional)
Short assignment, done at home, plenty of time
2nd interview (90 min)
With both Heads or the CEO
Reference check
Only people you've agreed to
Offer
We know that putting together a full resume and cover letter is a hassle. So we ask for just one thing: a 1-page CV. Please include the personal or project story you love most. We'd rather meet you on a relaxed Google Meet for everything else.
Attach your CV as a PDF. If anything goes wrong, email career@conscience.technology directly.