Akshant Jain

AI-native data and agents platform: warehouse, ETL, lineage, BI, multi-agent orchestration and governance. I own and lead its BI analytics module and its AI agent.

• Own the BI analytics module end to end and lead 2 backend engineers, 1 AI engineer and 1–3 QA testers; responsible for its system design (query layer, object-oriented design, data structures, latency) and its AI agent.
• Work with product and stakeholders to turn business needs into specs; built data querying over JDBC and Spark and a search interface that lets non-technical users explore their data.
• Built the first AI features on OpenAI models (natural-language charts, insights and suggestions), cutting users' manual analysis effort by ~90%.
• Rebuilt the BI agent as a multi-layer agent with transformation, classification, visualization and configuration layers; the AI team adopted this layered design for other complex agents.
• Own DAVE, DataGOL's AI analytics agent: it writes queries on selected tables, finds patterns and anomalies, builds charts, widgets and reports, then reads the results back to check its own insights.
• Brought back twice to take over the agent after AI-team versions fell behind the BI module; raised DAVE's evaluation scores from 2–3 to 4–5 (out of 5), cut latency ~50%, and cut errors and invalid configs by ~2/3.
• Restructured DAVE's skills, prompts and workflows with a config validation loop and tests, keeping token use in check and bringing the agent to parity with BI widget features.
• Built an evaluation framework that scores the agent's output with Playwright UI checks of rendered charts and widgets plus content checks.
• Took over a client's AI agent (grocery app) on a 2-week deadline: fixed its pipelines and logic for 2–3x lower latency, restructured prompts to cut size and cost, and advised on next features like memory and alerts.
• Designed a transformation layer on top of the BI module so manual and AI-generated analytics share one format, letting new structures be added without changing BI internals.
• Architected the dashboard engine (drill-down, conditional formatting, optimized SQL, in-memory caching, parallel queries) for sub-second responses on large datasets.
• Built a white-label dashboard embedding framework with theming and API-based configuration, a reporting engine with ad-hoc queries and PDF/Excel/CSV export, and event-driven threshold alerts.

Venture-backed SaaS deal platform connecting sponsors, lenders and advisors (client platform). Since 2024 my role has been lead and mentor as my focus moved to DataGOL.

• Helped build and scale the platform, which processes $50B+ in transactions for 600+ institutional clients at 99.9% uptime, through its venture funding with Asymmetric Capital Partners.
• Built an AI document feature that turns uploaded financial documents into structured financial data and lets users ask plain-English questions about the extracted results.
• Migrated the platform from AngularJS to React, cutting API calls by 65% and page load time by 55%.
• Built an NDA security layer so sponsors can flag confidential deal information to lenders, and a coverage module that keeps the full history of lender–sponsor interactions.
• Built SOC 2-compliant multi-tenant collaboration tools (encryption, RBAC, audit trails) and REST microservices handling 1M+ daily API calls at 99.95% reliability.
• Served on the production monitoring rotation, fixing ~10 production bugs and shipping ~5 patches per month.

Material management for an additive manufacturing platform (client platform).

• Improved material control, planning and flow across manufacturing facilities, cutting errors by 50%.
• Integrated material management into the platform with automated calculations, cutting manual effort by 95% and speeding up the process by 25%.
• Built a library of reusable React components for the platform's new React front-end framework.
• Owned delivery end to end: requirements gathering, specs, coding, testing, deployment, user training and ongoing support.

Open-source packages (visvoai-ai, visvoai-core)

• visvoai-ai: a provider layer over LangChain chat models with a live model registry (pricing, context window, capabilities), cost metering and one set of reasoning levels across Anthropic, Gemini and OpenAI-compatible APIs.
• visvoai-core: a LangGraph agent runtime with override hooks for nodes, routing, state, checkpointing and persistence, so products extend it without forking, plus a step cap that forces a clean final answer.
• A single tool contract for typed functions, class-based tools and LangChain tools, with self-registration, validation at import and pluggable call logging.
• Hand-built BM25 + cosine tool search, generalized from the platform into the core package.
• CI enforces the boundary between the public packages and the private platform, so private code or names can't reach a public release.

Platform

• Agent harness and loop on LangGraph: a custom tool-execution node with parallel batching, repeat-call blocking, schema-guided recovery from bad arguments, step caps and context compaction.
• Dynamic tool binding so agents stop loading every tool, which had used ~57% of each prompt: per-step retrieval over Gemini-embedded MCP tools, and tool descriptions rewritten to lift recall@8 from 91% to 97% on a 211-case benchmark.
• A tool system of ~40 tools with roles and per-tool permission checks before execution, plus human-in-the-loop approvals (approve, edit, approve-all, reject) that pause and resume the agent.
• Custom agents users define at runtime, 17 on-demand skills, and subagent calling with isolated context, depth caps and parallel runs that stream live to the parent's UI.
• Versioned artifacts (docs, code, slides, sheets, charts) that agents and users co-edit with conflict checks, SQL over spreadsheets in DuckDB, and a network-isolated sandbox where agents run code.
• RAG over a per-user document Library: PDF/DOCX/PPTX parsing, heading-aware chunking, Gemini embeddings and hybrid BM25 + vector search in Weaviate with citations, plus long-term chat memory.
• An MCP client and OAuth connectors written from scratch (Google, Microsoft, Slack, Notion, HubSpot) that sync server tools by content hash, repair malformed LLM arguments and move large results to files.
• LLM cost and token tracking with a trace viewer, automatic audit logs, encrypted per-tenant API keys, conversation branching, and stream resume on page reload over Redis Streams.