Software Engineer, AI · Princeton, NJ · Open to remote work or relocation

Akshant Jain

I build AI agents end to end, from design and evaluation to agents running in production.

I lead BI analytics at DataGOL, where I built the analytics engine end to end and overhauled DAVE, its flagship AI agent, using an evaluation system I designed. Employee of the Year 2024.

  • 5 years of experience
  • Java
  • TypeScript
  • Python
  • LangGraph
  • Agent evals
  • MCP
View ResumeSchedule a Call

Impact

4
person BI team I lead

Lead DataGOL's 4-person BI team; BI is the platform's most-used feature.

~15s
DAVE answer time

Cut DAVE's response time on simple data questions from 45–60 to about 15 seconds, and lifted its answer quality from 2–3 to 4.1/5.

55–60%
faster client AI assistant

Cut an AI shopping assistant's response times by ~55–60% in a two-week POC, helping secure a 2-year client contract.

~10
parallel agent tasks with Repolane

Built Repolane, an open-source tool that runs about 10 AI coding agent tasks in parallel across 5–6 repositories, each in its own isolated worktree.

About

Portrait of Akshant Jain
Akshant Jain
Software Engineer, AI
Open to opportunities

I build AI systems end to end: the agent itself, the evaluations that show whether a change made it better, and the product and data layers around it. Much of what I know about this comes from making DAVE, an AI agent, reliable enough for clients to depend on: evaluating it against real client questions, identifying where it failed, and improving its speed without compromising answer quality.

I work evaluation-first. Before changing a prompt, a tool, or a model, I want a test that would catch a regression. That same approach shapes Repolane, my open-source tool for managing Claude Code work across multiple repositories, where guardrails are backed by regression tests.

I started as a front-end engineer and grew into full-stack and AI engineering over five years. I studied Computer Science with a minor in Mathematics at the University of Illinois Chicago.

Writing

Experience

  1. Software Engineer · Programmer Analyst until Oct 2023

    InnCreTech LLC
    Oct 2021 — Present

    DataGOL

    Oct 2023 — Present

    AI-native data and agents platform. I'm the BI Analytics Lead.

    • Launched DataGOL's first AI analytics agent, turning plain-English questions into visualizations and reducing new clients' reliance on training and manual analysis.
    • Overhauled its successor, DAVE (Data Analytics & Visualization Engineer), now DataGOL's flagship proprietary agent, lifting answer quality on general use cases from 2–3 to 4.1/5. Kept tuning it on real-user edge cases, re-running evals after each change.
    • Cut DAVE's response time on simple data questions from 45–60 to about 15 seconds via context engineering of its prompts, skills and tool calls.
    • Implemented a self-correcting feedback loop with tool-call validation, so DAVE catches and fixes visualization errors in milliseconds before users see them.
    • Created an LLM-as-a-judge eval pipeline with a golden dataset of real client questions, scoring DAVE on 11 criteria before every release to catch regressions.
    • Benchmarked DAVE across 4 LLMs on answer quality, latency and cost, and wrote a report to help clients choose which model to run.
    • Cut AI shopping assistant response times by ~55–60% (12–16s to 5–7s) by optimizing the agent harness during a two-week POC, helping secure a 2-year client contract.
    • Engineered the Java query layer behind DataGOL's BI module, generating dialect-specific SQL for 4 databases and cutting new-database integration from 1 week to 1–2 days.
    • Designed one visualization format (charts, tables, pivots, KPIs) shared by the no-code builder and AI agent, so new features ship to both in days.
    • Retained a key client by delivering Tableau-style advanced calculations (LOD expressions) on a tight deadline, and added cohort analysis along the way.
    • Lead a 4-person BI team, run client demos, and trained a teammate who now leads BI testing; BI is the platform's most-used feature.

    TermGrid

    Debt financing platform for private capital, used by 1,600+ institutions and 30,000+ professionals.

    • Prototyping an AI agent that reads financial documents and fills in the data automatically.
    • Built the Relationship Management Dashboard, bringing platform-wide data on each lender or sponsor firm into one view; it's widely used and helped win clients.
    • Migrated the data room (secure document sharing for deals) to React, planned easier navigation for deeply nested folders, and designed its NDA and encryption features.
    • Implemented the interface for Precedents, which lets users compare terms across deals and term-sheet versions, and partnered with the backend team on its design.
    • Rewrote the loan allocation and fee screens to make them much faster.
    • Collaborated on the AngularJS-to-React migration from day one, delivering twice my assigned share within 3 months; pages loaded 55% faster with 65% fewer server calls.

    CO-AM (formerly Link3D)

    Jan 2022 — Jan 2024

    Additive manufacturing platform. Link3D was acquired by Materialise for $33.5M.

    • Built material management end to end for 3D print jobs: tracks stock and batches, forecasts material for scheduled jobs, and alerts when stock runs low.
    • Developed, with two teammates, the React component library for moving the platform from Angular to React, required as part of the Materialise acquisition.

  2. Software Developer (Part-time)

    Stealth Startup
    Jul 2021 — Sep 2021
    • Designed and built the startup's customer-facing landing website for its product launch, plus front-end prototypes of popular existing sites for its testing phase.

  3. Web Developer (Intern) · New Delhi, India

    SVS Techno Consultants Pvt. Ltd.
    Jun 2020 — Aug 2020
    • Turned the company's website into a React and MySQL dashboard with Firebase authentication and two-step sign-in, improving progress updates between employers and administration by 25%.
    • Built and tested a Node.js email API with SendGrid, optimized performance with a caching system, and wrote testing scripts, working in the web team's Scrum workflow.

Projects

Sep 2026

Repolane

Problem
A single software task often spans several Git repositories, which makes context, branches and progress hard to manage, and AI coding agents easily stray into other repos, secrets or main branches.
Solution
An open-source (Apache 2.0) tool that groups the repositories a task needs into one lane of Git worktrees, so Claude Code sessions stay scoped to the task, share its progress, and run side by side.
WebsiteGitHub
PythonBashGitClaude CodeAI agents
The Repolane board: three active lanes and one parked, across five repos, with per-repo uncommitted and unpushed counts.The Repolane board: three active lanes and one parked, across five repos, with per-repo uncommitted and unpushed counts.
  1. 01LanesOne task, a Git worktree per repo
  2. 02ContextResume note, log and repos in one file
  3. 03GuardClaude Code hooks check supported tool calls
  4. 04Checks184 regression cases, 9 live probes
  • Designed a Git-worktree-based workspace model that groups multiple repositories under one task, enabling concurrent, task-scoped Claude Code sessions with shared progress context.
  • Built a Python rules engine on Claude Code hooks that inspects tool calls and denies disallowed operations, including access to recognized secret files, unrelated task workspaces and protected branches.
Read the full case study

Earlier projects

Show 6 college and early projectsHide

Cloudgile

Jan — May 2021

Agile project management web app built by a four-person team using Scrum: issues, sprints, backlog, chat and project history, with users and data allocated by project role.

ReactMaterial UIFirebaseJavaScript

My Daily Life

Aug — Dec 2020

Working prototype of a planner that brings a daily schedule, goals and checklists into one interactive overview, designed on UI principles in a five-person team.

ReactHTMLCSSNetlifyUI design

Recognition & education

Award · 2024

Employee of the Year

InnCreTech LLC

For leading the BI analytics module and helping win new clients.

Education · 2021

University of Illinois at Chicago

BS Computer Science, Mathematics minor

Dean's List (Fall 2019, Spring 2021) · ACM member (ACM UIC, SIG-WebDev)

Certifications
Show 3 Google Cloud certificationsHide

Recommendations from colleagues are on LinkedIn

Skills

Agent toolset

AI & agents
Agent designContext engineeringTool callingSelf-correcting feedback loopsGuardrailsAgent evals (LLM-as-a-judge, golden datasets, regression testing)LangGraphLangSmithLangChainMCPClaude Code hooksRAG

Engineering foundation

Languages
Java · TypeScript · Python · JavaScript · SQL · Bash
Frontend & data
React · Next.js · AntD · Data visualization and dashboards · Spark · BigQuery · Athena · Snowflake · PostgreSQL · MySQL · Redis
Tools & infrastructure
Playwright · Docker · GitHub Actions · GCP
Portrait of Akshant Jain
Open to opportunities

Building AI agents? Let's talk.

I'm looking for AI engineer roles where I can build agent systems end to end. Book a call or grab my resume, and I'll get back to you quickly.

View ResumeSchedule a Call