ABOUT THE CLIENT
Client Profile
Kalmantic, an AI native technology company building autonomous agentic systems that help engineering and research teams track and act on fast moving AI/ML research
Headquarters
United States, with a distributed engineering team
Specialization
Agentic AI systems for automated research analysis, trend detection, and benchmark evaluation across the AI/ML landscape
GOALS
- Eliminate manual, time intensive triage of newly published AI research papers across two production agentic systems
- Build reliable, automated analysis capability so the research team could focus on judgment calls rather than repetitive review
- Surface emerging trends across the AI/ML research landscape
- Automate evaluation benchmarking so new techniques could be validated consistently, without manual setup for every run
- Integrate advanced AI reasoning into both existing agentic systems as a natural extension, not a disconnected bolt on
- Move from concept to a live, production grade capability supporting daily research workflows
- Establish an ongoing support model to sustain and evolve the integrations as research priorities changed
CHALLENGES
Kalmantic's research team faced a research field moving faster than any single team could read. New AI/ML papers appeared daily, promising techniques needed to be weighed against real benchmarks, and separating genuine progress from noise consumed hours of manual analyst time each week. Doing this by hand did not scale, and the client's two proprietary agentic systems, built to help track the research landscape, lacked native AI reasoning integrated into their pipelines, leaving trend detection and benchmark validation as largely manual, engineering intensive processes.
The engagement demanded more than a bolt on model call. Hexaview's engineers needed deep, hands on familiarity with both existing agentic architectures before any integration could be designed, so that new AI capability would function as a natural extension of how each system already operated. Coordinating research grade accuracy requirements, a live production timeline, and the client's continuously evolving research priorities added further complexity to an already fast moving engagement.
SOLUTIONS
- Agentic Architecture Immersion: Hexaview's consulting and forward deployed engineers embedded directly with Kalmantic’s engineering team, building deep familiarity with both proprietary agentic systems before designing any integration
- AI Model Integration Design: Designed and implemented Claude-based integrations across both agentic systems, built as a natural extension of each system rather than a bolt-on
- Automated Paper Analysis: Delivered automated workflows that continuously analyze newly published AI research papers as they appear, removing manual review from the team's daily workload
- Trend Detection Capability: Implemented logic to surface emerging trends across the AI/ML research landscape as they develop, rather than relying on periodic manual scans
- Automated Benchmark Evaluation: Built evaluation benchmarking workflows that run consistently in the background, validating new techniques without manual setup for each cycle
- Production Rollout: Took both integrations live in production, embedding AI driven analysis directly into the client's daily research operations
- Ongoing Support Model: Established a continuing engagement to address new needs and issues as the client's research workflows and priorities evolve
IMPACTS
- Faster Research Triage: Cut time spent on manual research paper triage by approximately 75 %, freeing analysts to focus on judgment calls rather than manual review
- Higher Research Throughput: Increased the volume of papers and benchmarks processed each month by roughly 30 %
- Rapid Path to Production: Moved from initial engagement to a live, production grade capability within months, with both integrations operating in production since March 2026
- Earlier Trend Visibility: Emerging techniques and technology shifts across the AI/ML landscape now surface earlier, giving the research team a head start on evaluating promising approaches
- Consistent Benchmark Validation: Evaluation benchmarks now run automatically and consistently in the background, eliminating manual setup previously required for every validation cycle
- Natural Architectural Fit: New AI capability was delivered as a natural extension of both existing agentic systems, preserving system coherence for the client's own engineering team
- Sustained Engagement: An ongoing support relationship keeps the integrations evolving as Kalmantic's research priorities and system needs change over time
SUCCESS FACTORS
- Deep technical immersion in Kalmantic’s proprietary agentic architecture before any integration was designed
- A combined consulting and forward deployed engineering model that enabled hands on collaboration directly with the client's engineering team
- Architecture first integration approach that treated new AI capability as a natural extension of existing systems, not an isolated add on
- Focus on production grade reliability for research and benchmarking workflows that inform real technical decisions
- Structured transition from delivery into ongoing support, sustaining the relationship as client needs evolved
- Continuous, direct collaboration between Hexaview and client engineering teams throughout design, build, and rollout
"Claude has become core to how our research team works. Instead of spending hours manually going through new papers and setting up benchmarks by hand, our agents surface the trends that matter and run evaluations consistently in the background. Hexaview's team understood our systems well enough to build this in as a natural extension of OpenClaw and NemoClaw, not a bolt-on."
Kashi KS, Co-founder and CTO, Kalmantic
.svg)
.svg)