CareerTrack

Many employees do great work but don’t document it due to lack of a standard system. My goal is to help employees capture achievements early, making appraisals fair and effortless.

Comment

Inspiration

Many employees do excellent work throughout the year, but their efforts often go undocumented due to the lack of a simple, standard tracking system. This leads to incomplete or unfair appraisal discussions. CareerTrack was inspired by the need to help individuals consistently capture their achievements and make performance reviews stress-free and accurate.

What it does

CareerTrack helps employees log their work, achievements, challenges solved, and learnings throughout the year. It creates a structured, searchable record that can be easily summarized during appraisal cycles, ensuring no contribution is missed.

How we built it

I designed CareerTrack with a user-first approach, focusing on simplicity and consistency. The system uses structured inputs for achievements and challenges, backed by automation to organize entries and generate appraisal-ready summaries.

Challenges we ran into

  • Ensuring agents interpreted real user logs and signals, not generic or random prompts, while preserving context and intent.
  • Designing thought chaining to reliably link evidence, merge related entries, and maintain long-term continuity.
  • Determining the right number of tools and triggers to synchronize capture, refinement, and appraisal without over-engineering.
  • Balancing autonomous self-correction with minimal user interruptions.
  • Running extensive test cases to validate that the agent consistently captured accurate details, avoided hallucinations, and correctly understood its role across scenarios.
  • Finalizing guardrails

Accomplishments that we're proud of

  • Built a Gemini-powered autonomous agent that captures and refines career data over time with minimal user input.
  • Designed a non–prompt-based system using multi-step orchestration, long-term memory, and self-correction loops.
  • Successfully implemented thought chaining to link evidence, merge related work, and maintain continuity across entries.
  • Enabled appraisal-ready summaries that are fully traceable to documented work, preventing hallucinations.
  • Created a clean, judge-friendly UI that demonstrates Gemini’s reasoning and autonomy within a short demo window.

What we learned

  • Less input, better design: Users are more consistent when they can log work in seconds, letting the agent handle structure and refinement.
  • Autonomy beats interaction: Long-running agents with background reasoning provide more value than prompt-driven experiences.
  • Confidence scoring is critical: Explicit uncertainty handling helps avoid hallucinations and guides when clarification is truly needed.
  • Memory needs structure: Thought signatures and chaining are essential for merging related work and maintaining long-term context.
  • Testing defines trust: Running diverse real-world scenarios is key to ensuring the agent understands its role and produces reliable outputs.

What's next for CareerTrack

  • Add AI-generated goal alignment to map daily work against role expectations and performance frameworks.
  • Introduce manager and reviewer views for shared visibility and feedback without manual reporting.
  • Expand evidence ingestion (PRs, tickets, docs) with deeper reasoning over artifacts.
  • Enable cross-cycle comparisons to track growth and impact over multiple appraisal periods.
  • Integrate with work tools (Git, Jira, Slack) to capture signals automatically while keeping user effort minimal.

Built With

Share this project:

Updates