Senior Software Engineer, AI Research

kAIgentic; Bengaluru; On-site; INR 5,500,000 - 9,000,000; Full time

Job summary

kAIgentic is building the intelligence layer for enterprises, delivering a governed ontology and evidence-backed outputs to improve decision making. The role demands deep AI research expertise in regulated settings, with strong Python/PyTorch skills and mastery of graph databases.

You will own end-to-end design of knowledge graphs, extraction pipelines, and RL-based evaluation, contributing to production-grade retrieval-generation and scalable learning loops across cross-functional teams.

Qualifications

• PhD in AI or Computer Science or related field with at least 4 years of industry research experience.
• Proven expertise delivering end-to-end AI research systems in regulated settings.
• Deep mastery of Python, PyTorch, and model fine-tuning techniques (SFT, DPO, RFT).
• Strong knowledge of graph databases and ontology engineering, including provenance tracking.

Responsibilities

• Own end-to-end design and implementation of the knowledge-graph schema, including hierarchy, provenance, versioning, and extensibility.
• Build grounded-extraction pipelines that attach source evidence and epistemic status to every claim.
• Design and improve multi-context harnesses that manage memory tiers, budget constraints, and compaction for large-scale inference.
• Deliver cross-boundary knowledge distillation workflows that integrate external data sources into the internal ontology.
• Own consolidation frameworks that merge heterogeneous documents, interviews, and event logs into a unified model with conflict resolution.
• Build and iterate model adaptation loops, applying SFT, DPO, and RFT to balance engineering context and model capabilities.
• Design learning-loop infrastructure that captures feedback, drives active learning, and updates the ontology continuously.
• Own RL-environment and curriculum components, defining reward signals and evaluation metrics for safe operation.
• Build production-grade retrieval-generation pipelines that meet quality, latency, and cost budgets.
• Deliver comprehensive evaluation frameworks with blind rubrics and LLM-as-judge validation to measure grounding and hallucination.

Skills

AI research systems

Python

PyTorch

SFT

DPO

RFT

Graph databases

Ontology engineering

RL environments

RAG

Education

PhD in AI or CS

Tools

Ontology design tools

RAG pipelines

Model fine-tuning

Job description

kAIgentic is building the intelligence layer for the world's most ambitious enterprises. Headquartered in Singapore with teams in India and Japan, our software platform helps large organizations evolve as fast as technology itself by turning the tacit know-how locked inside their people into safe, governed, AI-powered operations.

The hardest part of enterprise transformation is not strategy. It is execution. Institutional knowledge lives in people's heads, systems are fragmented, and risk tolerance is low. kAIgentic captures how work actually happens, designs better workflows, and runs them inside an intelligence layer that is observable, auditable, and engineered for the most regulated environments on earth. The outcome is an enterprise that continuously improves.

We are backed by SMBC Group as our founding partner and customer zero, and our platform is already being proven inside one of the most complex, regulated operating environments in the world. That means real problems, real data, and real production impact from Day 1.

The Role

Own the intelligence layer that transforms enterprise process knowledge, documents, interviews, system logs, into a governed, provenance-rich ontology, and the extraction and learning systems that consume it. Deliver evidence-backed outputs; any hallucination is a product failure.

What You’ll Do

• Own end-to-end design and implementation of the knowledge-graph schema, including hierarchy, provenance, versioning, and extensibility.
• Build grounded-extraction pipelines that attach source evidence and epistemic status to every claim.
• Design and improve multi-context harnesses that manage memory tiers, budget constraints, and compaction for large-scale inference.
• Deliver cross-boundary knowledge distillation workflows that integrate external data sources into the internal ontology.
• Own consolidation frameworks that merge heterogeneous documents, interviews, and event logs into a unified model with conflict resolution.
• Build and iterate model adaptation loops, applying SFT, DPO, and RFT to balance engineering context and model capabilities.
• Design learning-loop infrastructure that captures feedback, drives active learning, and updates the ontology continuously.
• Own RL-environment and curriculum components, defining reward signals and evaluation metrics for safe operation.
• Build production-grade retrieval-generation pipelines that meet quality, latency, and cost budgets.
• Deliver comprehensive evaluation frameworks with blind rubrics and LLM-as-judge validation to measure grounding and hallucination.

What You’ll Bring

• Proven expertise delivering end-to-end AI research systems from prototype to production in regulated settings.
• AI-native velocity as a default mode of working
• Deep mastery of Python, PyTorch, and large-scale model fine-tuning techniques such as SFT, DPO, and RFT.
• Expert knowledge of graph databases and ontology engineering, including schema design, version control, and provenance tracking.
• Strong experience building extraction pipelines that link claims to source evidence and manage epistemic status.
• Hands‑on skill with RL environments, curriculum design, and reward modeling for autonomous systems.
• Proficiency in retrieval-augmented generation, latency optimization, and cost-aware scaling.
• Ability to design and run rigorous applied-research experiments, including hypothesis formulation, ablation studies, and statistical validation.
• Excellent judgment to balance model adaptation versus context engineering, and clear communication with cross‑functional stakeholders.
• PhD in Artificial Intelligence or Computer Science, or a related field, with at least 4 years of relevant industry experience in Research

Why join kAIgentic?

We are a global team of builders who thrive in ambiguity, care deeply about the customers we serve, and believe the intelligence layer is how enterprise work will be reshaped over the next decade. We are building the connective tissue that lets large companies operate with the speed of a startup and the trust of an institution.

We look for people who:

• Combine technical excellence with genuine customer empathy.
• Are entrepreneurial and energized by zero-to-one problems with no playbook.
• Lead with ownership, integrity, and collaboration, not titles.
• Want to help define a new category of enterprise AI, not just ship inside an existing one.
• Working here means being surrounded by peers who challenge assumptions, celebrate progress, and build with both courage and care.

Life at kAIgentic

• Intelligence layer at the core. You will be shaping the substrate that turns institutional knowledge into governed, production-grade operations. This is enterprise infrastructure with real consequences.
• Innovation at enterprise scale. Startup velocity meets the depth, scale, and stakes of mission-critical, regulated environments. Both are non-negotiable.
• Ownership from Day One. Your work directly shapes the product, the culture, and the outcomes our customers see.
• Learning and growth. You will work alongside seasoned leaders from leading enterprises who have built and scaled global businesses.
• A culture of trust. Psychological safety, transparent disagreement, and disciplined experimentation are how we operate, not slogans on a wall.
• Global collaboration. Teams across Singapore, India, Japan, Europe, and the US, working as one.
• A mission worth the effort. Building something the world has not seen before: anintelligence layer that helps enterprises continuously improve how they run.