Apeksha Awasthi

Current truth

• Product manager for the MCP team at lucid (aawasthi@lucidchart.com). Our main point of contact for any MCP server changes / tool updates on the ChatGPT-app eval engagement — she routes the changes we want to test into the test server (with ryan-ferguson as the developer).
• Based on the US East Coast — works from Lucid’s Raleigh, North Carolina office (~150 people there), not the Salt Lake HQ.
• Owns the baseline-review homework: re-classify partials/fails per the agreed methodology and send a prioritized critical-issues list with repro prompts (async, before the ~Tue 2026-07-07 sync).
• [idea] Her team’s focus pivots next quarter from tool-building (RunScript absorbs that) to enterprise admin controls/governance: IP allowlisting, client connect restrictions, MCP audit logging — her own read of her roadmap, 2026-07-01.
• What she cares about in evals: limitation honesty — the model should say “I can’t do that” upfront instead of attempting workarounds. Constraint she enforces: Lucid enterprise policy prohibits storing/viewing user prompts, even anonymized; her insight substitute is 2–4 user interviews/week.
• Working style read (ours, from stand-ups): her pace reflects Lucid’s org speed, not disinterest.

Open questions

• Name spelling inferred from email (aawasthi) + audio; rendered “Apeksha”/“Abhisha”/“Apexha”/“Apexa”/“Apecha” across transcripts. Confirm exact spelling.
• Seniority and who she reports to.

Timeline

• [2026-07-02] (call) Attended the bi-weekly sync (joined late). Confirmed no recent Claude-side listing changes despite the discoverability score climb. Took two async items for ~Tue Jul 7: baseline review (edit sheet or bullets) + prioritized issues with repro prompts; availability limited by Jul 4 + hackathon week.
• [2026-07-01] (call) 1:1 with Vincent: aligned eval methodology (test today’s capabilities; happy path/edge/negative), reclassified sample partials/fails, disclosed RunScript and her team’s pivot to enterprise governance, ruled out prompt-keyword capture on enterprise-policy grounds, described the Codex-plugin skills process (written in ~a week from known flows; expects it needs to get more robust).
• [2026-06-18] (call) Attended the bi-weekly sync; confirmed as PM for the MCP team and the point of contact for MCP server iteration. Took ownership of reviewing the eval partials.